rentahuman-mcp 3.1.0 → 3.2.0

package/README.md

@@ -105,8 +105,9 @@ webhooks.
 
 ### Conversations
 
-- **start_conversation**: Start a conversation with a human when the account is eligible for direct messaging. Prefer `create_bounty` for first contact.
-- **send_message**: Send messages in an existing conversation. Accepts optional `idempotencyKey`.
+Direct messaging is no longer available. Use `create_bounty` for first contact. Existing direct threads remain readable but read-only; messaging is available for bounty, service-booking, feedback, dispute, and platform-support conversations. `start_conversation` is deprecated and hidden from the tool catalog, and cached calls return an error directing clients to `create_bounty`.
+
+- **send_message**: Send messages in an existing supported conversation. Accepts optional `idempotencyKey`.
 - **get_conversation**: Get conversation history with all messages
 - **list_conversations**: List all your conversations with humans
 
@@ -193,10 +194,10 @@ worker (`escrow_recipient_mismatch`).
 
 ### Bounties (Task Postings)
 
-- **create_bounty**: Post a one-shot task bounty for humans to apply to (supports multi-person and country-wide bounties). Accepts optional `idempotencyKey`. Dry-run `preview.fundingTotal` is the total funding requirement before existing wallet balance is applied. Live `pending_deposit` `checkout_total` is the exact hosted charge after wallet balance is applied. Ongoing data-collection bounties (`bountyKind: "ongoing"`) are Figure ongoing-operator-only REST resources and are intentionally not exposed through MCP.
+- **create_bounty**: Post a one-shot task bounty for humans to apply to (supports multi-person and country-wide bounties). Accepts optional `idempotencyKey`. Dry-run `preview.fundingTotal` is the total funding requirement before existing wallet balance is applied. Live `pending_deposit` `checkout_total` is the exact hosted charge after wallet balance is applied.
 - **list_bounties**: Browse available bounties (includes partially filled bounties by default)
 - **get_bounty**: Get detailed bounty information including spots filled/remaining
-- **update_bounty**: Modify an ordinary one-shot bounty or close it while unassigned. Work Completed and Paid are synchronized from escrow evidence and cannot be set directly. Figure ongoing-operator-only bounty settings are intentionally omitted from MCP.
+- **update_bounty**: Modify an ordinary one-shot bounty or close it while unassigned. Work Completed and Paid are synchronized from escrow evidence and cannot be set directly.
 - **cancel_bounty**: Cancel one of your bounties by ID with a required canonical reason and `details` only when the reason is `other`; this feedback is private and never shown to workers
 - **get_bounty_applications**: View applications for your bounty, including uploaded media URLs (`imageUrls`, `videoUrls`, `documentUrls`) on upload-collection bounties. Each application carries `isBlockedByOwner` and `isPreferredByOwner`; call `prefer_human` / `block_human` with the application's `humanId` to favorite or block an applicant while reviewing (blocking is future-only and never removes an accepted worker). Each application also carries `reviewSummary` (`averageRating` 1-5 or `null`, `reviewCount`, up to 3 `recentReviews` with `rating`, `comment`, `escrowVerified`, `reviewerName`, `createdAt`); `averageRating: null` with `reviewCount: 0` means no review history yet, not a low score
 - **get_bounty_dataset**: List download URLs for every accepted upload on one of your upload-collection bounties
@@ -242,14 +243,15 @@ authorize payment.
 
 ### Human-written text
 
-- **create_humanization**: Hire one human for a private text rewrite with a fixed worker price and exact turnaround. Requires a stable `idempotencyKey`. Optional applicant screening takes caller-supplied AI-generated text, sizes it by percentage with a 500-word cap, and requires the same document, continuous recording, attestation, and advisory Pangram analysis before acceptance. A private resume upload can also be required.
-- **get_humanization**: Poll the owned order, applicant screening evidence/results, current worker deadline, attempts, submission links, review state, and advisory Pangram results.
+- **create_humanization**: Hire one human for a private text rewrite with a fixed worker price and exact turnaround. Requires a stable `idempotencyKey`. Optional applicant screening takes caller-supplied AI-generated text, sizes it by percentage with a 500-word cap, and requires the same document, continuous recording, attestation, and Pangram analysis before acceptance. A valid AI fraction below 50% approves the screening; 50% or greater rejects it. Missing or invalid scores remain in manual review. A private resume upload can also be required.
+- **get_humanization**: Poll the owned order, applicant screening evidence/results and automatic review state, current worker deadline, attempts, submission links, final review state, and Pangram results.
 
 Humanization source text and private instructions are never included in the
 public bounty. Detector scores and screen recordings are review evidence, not
-proof of authorship; the tools never average scores or make an automatic
-payment/rejection decision. Use the existing submission-review tool to approve,
-request a redo, or reject after analysis becomes terminal. Requests that seek
+proof of authorship. The platform applies the 50% AI-fraction threshold to
+screening samples and managed delivery review, but it never releases payment by
+score alone. Use the existing submission-review tool for any manual approve,
+redo, or reject decision after analysis becomes terminal. Requests that seek
 credential falsification, recommender impersonation, academic-integrity
 violations, or required-disclosure evasion are rejected.
 
@@ -272,6 +274,20 @@ payout is confirmed. Agents cannot set `completed` or `paid` through
 
 `create_bounty` and `update_bounty` accept `identityRequired` (boolean, default false). When true, applicants must pass an identity check (government ID) before they can apply. Identity is verified once per account and reused across every bounty that requires it, so a returning worker who is already verified applies without re-checking. Recommended for in-person tasks.
 
+#### Excluding previous participants
+
+`create_bounty` and `update_bounty` accept `excludedParticipantSourceBountyIds`
+(up to 1,000 ids of your own earlier bounties). Any worker who was ever accepted
+on one of those bounties cannot see or apply to the bounty, which is how repeat
+surveys and fresh-panel studies keep prior participants out. `update_bounty`
+replaces the whole list: pass the full new list, or an empty array to clear it.
+Exclusion is evaluated live, so a source added later hides the bounty from its
+ever-accepted workers and blocks accepting them; pending applications are not
+auto-rejected. Sources must be your own regular bounties that were funded at
+least once, and never the bounty itself: AI-managed, QA, and taste run
+bounties, archived program records, never-funded drafts, and a self-reference are
+rejected with a 400 whose `error` names the problem.
+
 #### Country targeting
 
 `create_bounty` and `update_bounty` accept a country-only `location` when a
@@ -337,10 +353,13 @@ review and accept every application yourself. Ignored on `aiManaged` bounties,
 whose managed engine owns applicant review.
 
 On `micCheckRequired` bounties the deterministic gate scores the applicant's
-mic recording: auto-accept requires an overall DNSMOS score of at least
-`autoAcceptMinMicScore` (1-5, default 3.0 — the platform's pass boundary).
-Set it at creation or change it later via `update_bounty` (`null` resets to
-the default). Below the floor, applications wait for your manual review.
+mic recording by its overall DNSMOS score (1-5). Set `autoAcceptMinMicScore`
+to make that score a hard cutoff: applications below it are auto-rejected
+with a reason, and applications at or above it proceed toward acceptance
+(unscored recordings always wait for manual review). Set it at creation or
+change it later via `update_bounty` (`null` resets to the platform default of
+3.0, below which applications only wait for your manual review — nobody is
+rejected for audio quality by default).
 
 #### Microphone quality requirement
 
@@ -421,7 +440,20 @@ Create bounties that hire multiple humans (e.g., "10 people to hold signs in Tim
 - Bounty status transitions: `open` → `partially_filled` → `assigned` →
   `completed` (**Work Completed**) → `paid`; every accepted seat must satisfy
   the release/payout condition before the aggregate advances.
-- Other applications are only auto-rejected when all spots are filled
+- Recruiting continues while spots are merely filled (a seat can still be
+  released and refilled): new applications keep arriving as pending, existing
+  pending applications are kept, and platform outreach keeps running. The pool
+  is only auto-rejected — and recruiting only stops — once every seat's escrow
+  is settled (released/paid). Accepting more workers than funded seats stays
+  impossible.
+- Staffed bounties that are still recruiting appear in `list_bounties`'
+  default recruiting list (they carry `spotsRemaining: 0`). Pause one with
+  `update_bounty` (status `paused`) to stop applications and outreach while
+  accepted workers keep working; resuming with status `open` returns it to the
+  seat-derived status (`open`, `partially_filled`, or `assigned`).
+- With `autoAccept`, applicants who arrive while every seat is taken wait as
+  pending and are re-reviewed automatically, oldest first, whenever a seat is
+  released.
 
 ### Upload-Collection Bounties (photos, videos, documents)
 
@@ -451,10 +483,6 @@ Collect media or documents directly instead of standard applications — e.g. "u
 
 3. **Retrieve the dataset** of everything you accepted with `get_bounty_dataset(bountyId)` — returns download URLs for all accepted files grouped by applicant. For a bulk zip archive, `GET https://rentahuman.ai/api/bounties/{bountyId}/applications/dataset` with your `X-API-Key` header.
 
-### Ongoing Data-Collection Bounties
-
-Ongoing bounties are specialized partner programs created with `bountyKind: "ongoing"` and an `ongoing` configuration block. They require the server-derived Figure ongoing-operator capability in the REST API and are not agent-facing MCP tools. Human contributors join those programs through the web data-collection intake flow, not through `create_bounty` or `update_bounty`.
-
 ## Usage
 
 ### TypeScript Types

package/dist/serve.js

@@ -844,7 +844,7 @@ Use \`get_reviews\` to see what other agents say about a human:
 
 ## Bounties and Messaging
 
-Post bounties for first contact. Direct messaging is eligibility-gated and should not be treated as the default way to reach humans.
+Post bounties for first contact. Direct conversation creation is retired; existing legacy direct threads are readable but cannot receive new messages.
 
 ### Post a Bounty
 
@@ -924,14 +924,11 @@ describing your relationship with that applicant: \`isBlockedByOwner\` and
 
 ### Start a Conversation
 
-Use \`start_conversation\` only when the account is eligible for direct messaging:
-- Provide a clear subject line
-- Write a friendly, specific initial message
-- The human will see your message in their account
+\`start_conversation\` is retired and returns an error. Create a bounty first; after a human applies, use its conversation to coordinate the task.
 
 ### Continue the Discussion
 
-Use \`send_message\` to reply in a conversation:
+Use \`send_message\` to reply in a bounty, service-booking, feedback, dispute, or platform-support conversation. Legacy direct threads are read-only:
 - Reference the conversation ID
 - Check for replies with \`get_conversation\`
 
@@ -1325,15 +1322,10 @@ var agentCheckoutTools = [
   refundAgentCheckout
 ];
 
-// ../contracts/src/tool-specs/bounties.ts
-import { Schema as Schema8 } from "effect";
-
-// ../contracts/src/schemas/requests/bounty.ts
-import { Schema as Schema5 } from "effect";
-
 // ../contracts/src/constants.ts
 var BOUNTY_PRICE_TYPES = ["fixed", "hourly"];
 var BOUNTY_MIN_PRICE_USD = 3;
+var MAX_EXCLUDED_PARTICIPANT_SOURCE_BOUNTIES = 1e3;
 var HUMANIZATION_FORMATS = ["text"];
 var HUMANIZATION_TRANSFORMATIONS = [
   "paraphrase",
@@ -1476,7 +1468,11 @@ function calculatePlatformFee(baseAmountCents) {
   };
 }
 
+// ../contracts/src/tool-specs/bounties.ts
+import { Schema as Schema8 } from "effect";
+
 // ../contracts/src/schemas/requests/bounty.ts
+import { Schema as Schema5 } from "effect";
 var BountyLocationRequest = Schema5.Struct({
   city: Schema5.optional(Schema5.String.pipe(Schema5.maxLength(100))),
   state: Schema5.optional(Schema5.String.pipe(Schema5.maxLength(100))),
@@ -1653,7 +1649,7 @@ var CreateBountyRequest = Schema5.Struct({
     description: "USD (default and only supported bounty currency)"
   }),
   bountyKind: Schema5.optional(Schema5.Literal("one_shot")).annotations({
-    description: "Agent-facing bounty creation only supports one-shot bounties. Ongoing data-collection bounties use bountyKind='ongoing' but are admin-only through the REST API and are intentionally not exposed through MCP create_bounty."
+    description: "Agent-facing bounty creation only supports one-shot bounties."
   }),
   spotsAvailable: Schema5.optional(
     Schema5.Number.pipe(
@@ -1684,7 +1680,7 @@ var CreateBountyRequest = Schema5.Struct({
     description: "Auto-reassign completion deadline: hours a worker has to complete the task after confirming their seat. Any whole number of hours from 1 to 720 (30 days) is accepted, e.g. 36 for a day and a half; the web form exposes the same range through its custom duration option. Workers are reminded at half the window; when the deadline passes the seat is automatically released and the listing reopens for other applicants. Workers may request an extension, which you approve, deny, or replace with your own via POST /bounties/{id}/applications/{appId}/extension/decision. For AI-managed bounties this overrides the default 6-hour work window. Omit for no completion deadline \u2014 and always omit for tasks anchored to a specific future date or time slot (events, shifts, appointments): the clock starts at seat confirmation, so a window shorter than the wait would release every worker before the task could start."
   }),
   keepApplicantsOnFill: Schema5.optional(Schema5.Boolean).annotations({
-    description: "When true, pending applicants are kept (not auto-rejected) once all seats fill, so you can reuse the applicant pool after reopening or adding seats. Default false."
+    description: "When true, pending applicants are kept (not auto-rejected) even after every seat escrow settles, so you can reuse the applicant pool after reopening or adding seats. By default applicants are already kept while seats are merely filled and are only auto-rejected at settlement. Default false."
   }),
   autoAccept: Schema5.optional(Schema5.Boolean).annotations({
     description: "Automatically review applicants as they apply. Deterministic checks (account standing, country eligibility, payout viability, your blocklist) always run; an AI review runs only when the application contains free-text screening answers that need judgment. Qualified applicants are accepted (the standard worker-confirmation window still applies unless skipAcceptanceConfirmation is set), clear mismatches are rejected with a reason, and uncertain cases stay pending for your manual review. Default true; pass false to review and accept every application yourself. Ignored for aiManaged bounties, whose managed engine owns applicant review."
@@ -1695,11 +1691,18 @@ var CreateBountyRequest = Schema5.Struct({
       Schema5.lessThanOrEqualTo(5)
     )
   ).annotations({
-    description: "Minimum overall DNSMOS speech-quality score (1-5) an applicant\u2019s mic-check recording must reach for auto-accept on micCheckRequired bounties. Below the floor the application stays pending for your manual review (never auto-rejected). Default 3.0 \u2014 the platform\u2019s pass boundary. Only meaningful with micCheckRequired: true and autoAccept on."
+    description: "Minimum overall DNSMOS speech-quality score (1-5) an applicant\u2019s mic-check recording must reach on micCheckRequired bounties. Setting this makes it a hard cutoff for automatic review: applications scoring below it are auto-rejected with a reason, at/above proceed toward acceptance. Unscored recordings always stay pending. When omitted, the platform default 3.0 applies and below-default applications stay pending for your manual review instead of being rejected. Only meaningful with micCheckRequired: true and autoAccept on."
   }),
   identityRequired: Schema5.optional(Schema5.Boolean).annotations({
     description: "Require applicants to pass an identity check (government ID) before applying. Verified once per account and reused across bounties. Default false."
   }),
+  excludedParticipantSourceBountyIds: Schema5.optional(
+    Schema5.Array(Schema5.String).pipe(
+      Schema5.maxItems(MAX_EXCLUDED_PARTICIPANT_SOURCE_BOUNTIES)
+    )
+  ).annotations({
+    description: `Exclude previous participants: ids of your own earlier bounties (max ${MAX_EXCLUDED_PARTICIPANT_SOURCE_BOUNTIES}). Any worker who was ever accepted on one of them cannot see or apply to this bounty. Editable later with PATCH. Ids must belong to your own regular bounties (not AI-managed, QA, or taste runs).`
+  }),
   micCheckRequired: Schema5.optional(Schema5.Boolean).annotations({
     description: "Require applicants to pass an on-device microphone quality check (live recording scored for noise, clipping, Bluetooth-headset bandwidth, and DNSMOS speech quality) before applying. Applicants get up to 2 attempts; the result (pass/warn/fail plus scores) is visible to you on each application via get_bounty_applications. Default false."
   }),
@@ -1707,7 +1710,7 @@ var CreateBountyRequest = Schema5.Struct({
     description: "BETA \u2014 fully AI-managed bounty. The platform recruits and vets workers, reviews their chat-delivered submissions against the description and completion criteria, releases payment for approved work, refunds unused budget to your wallet, and returns a final report (poll get_bounty's managed run or subscribe to the run.report_ready webhook). Requires fixed price in USD and 1-50 spots. Currently in beta \u2014 the API returns 403 managed_bounties_beta_disabled when unavailable."
   }),
   asOrg: Schema5.optional(Schema5.Literal(true)).annotations({
-    description: "Post this bounty on behalf of your organization. Requires the authenticated account to be a member of an organization (created from account settings); the org is resolved server-side and its members can co-manage the bounty. Funding still comes from your own wallet. Returns 403 when you don't belong to an organization."
+    description: "Feature-gated: use only when organization settings are available for the authenticated account. Posts the bounty on behalf of that account's organization; the org is resolved server-side and its members can co-manage the bounty. Funding still comes from your own wallet. Returns 403 when organizations are unavailable or you don't belong to one."
   }),
   submissionMode: Schema5.optional(
     Schema5.Literal(...BOUNTY_SUBMISSION_MODES)
@@ -1897,6 +1900,13 @@ var UpdateBountyRequest = Schema5.Struct({
   identityRequired: Schema5.optional(Schema5.Boolean).annotations({
     description: "Require applicants to pass an identity check (government ID) before applying. Verified once per account and reused across bounties."
   }),
+  excludedParticipantSourceBountyIds: Schema5.optional(
+    Schema5.Array(Schema5.String).pipe(
+      Schema5.maxItems(MAX_EXCLUDED_PARTICIPANT_SOURCE_BOUNTIES)
+    )
+  ).annotations({
+    description: `Replace the exclude-previous-participants list: ids of your own earlier regular bounties (max ${MAX_EXCLUDED_PARTICIPANT_SOURCE_BOUNTIES}), never this bounty itself. An empty array clears it. Exclusion is evaluated live, so a source added later hides the bounty from its ever-accepted workers and blocks accepting them; pending applications are not auto-rejected.`
+  }),
   micCheckRequired: Schema5.optional(Schema5.Boolean).annotations({
     description: "Require applicants to pass an on-device microphone quality check before applying (max 2 attempts; result visible on each application)."
   }),
@@ -1935,7 +1945,7 @@ var UpdateBountyRequest = Schema5.Struct({
     description: "Acknowledges the wallet impact of a wallet-funded bounty edit. Any edit that debits the poster wallet (adding seats via spotsAvailable, or raising the price) is first refused with error_code wallet_impact_confirmation_required and the exact amount; retry the same request with this set to true to apply the charge. Admins editing a bounty they do not own must also pass it for any edit that debits or credits the poster wallet."
   }),
   keepApplicantsOnFill: Schema5.optional(Schema5.Boolean).annotations({
-    description: "When true, pending applicants are kept (not auto-rejected) once all seats fill, so you can draw from the same pool after reopening or adding seats. Default false."
+    description: "When true, pending applicants are kept (not auto-rejected) even after every seat escrow settles, so you can draw from the same pool after reopening or adding seats. By default applicants are already kept while seats are merely filled and are only auto-rejected at settlement. Default false."
   }),
   autoAccept: Schema5.optional(Schema5.Boolean).annotations({
     description: "Turn automatic applicant review on or off. When on, deterministic checks always run and an AI review runs only for free-text screening answers; qualified applicants are accepted, clear mismatches rejected, uncertain cases left pending. Applications received while off stay pending; toggling on affects future applications only. Ignored on aiManaged bounties."
@@ -1948,7 +1958,7 @@ var UpdateBountyRequest = Schema5.Struct({
       )
     )
   ).annotations({
-    description: "Change the minimum overall DNSMOS mic score (1-5) for auto-accept on micCheckRequired bounties, or pass null to reset to the platform default (3.0). Affects future applications only."
+    description: "Change the minimum overall DNSMOS mic score (1-5) for automatic review on micCheckRequired bounties \u2014 applications scoring below your configured cutoff are auto-rejected, at/above proceed toward acceptance \u2014 or pass null to reset to the platform default (3.0, below which applications only stay pending). Affects future applications only."
   }),
   completionCriteria: Schema5.optional(
     Schema5.String.pipe(Schema5.minLength(10), Schema5.maxLength(2e3))
@@ -2265,29 +2275,6 @@ var LifecycleMessagesResponse = Schema7.Struct({
   onSubmissionApproved: Schema7.optional(LifecycleMessageTemplateResponse),
   onSubmissionRejected: Schema7.optional(LifecycleMessageTemplateResponse)
 });
-var BountyApplicationStepResponse = Schema7.Literal(
-  "verify-recording-device",
-  "set-up-payments",
-  "collect-email",
-  "confirm-equipment",
-  "collect-shipping-address",
-  "instructions",
-  "acknowledge-agreements"
-);
-var BountyApplicationFlowResponse = Schema7.Struct({
-  steps: Schema7.Array(BountyApplicationStepResponse)
-});
-var PaymentsSetupCapabilityResponse = Schema7.Struct({
-  supportedRails: Schema7.Array(Schema7.Literal("stripe", "wise")).pipe(
-    Schema7.minItems(1)
-  )
-});
-var PartnerCapabilitiesResponse = Schema7.Struct({
-  verifyRecordingDevice: Schema7.optional(Schema7.Literal(true)),
-  gatherShippingAddress: Schema7.optional(Schema7.Literal(true)),
-  paymentsSetup: Schema7.optional(PaymentsSetupCapabilityResponse),
-  applicationFlow: Schema7.optional(BountyApplicationFlowResponse)
-});
 var BountyPublicResponse = Schema7.Struct({
   id: Schema7.String,
   title: Schema7.String,
@@ -2333,9 +2320,12 @@ var BountyPublicResponse = Schema7.Struct({
   autoAccept: Schema7.optional(Schema7.Boolean),
   autoAcceptMinMicScore: Schema7.optional(Schema7.Number),
   requiredQualificationIds: Schema7.optional(Schema7.Array(Schema7.String)),
+  /** Owner-only: source bounty ids for participant exclusion. Never sent to workers. */
+  excludedParticipantSourceBountyIds: Schema7.optional(
+    Schema7.Array(Schema7.String)
+  ),
   supportedCountries: Schema7.optional(Schema7.Array(Schema7.String)),
   intakeClosedCountries: Schema7.optional(Schema7.Array(Schema7.String)),
-  partnerCapabilities: Schema7.optional(PartnerCapabilitiesResponse),
   submissionMode: Schema7.optional(Schema7.String),
   qaSubmissionChannel: Schema7.optional(Schema7.Literal("chat_delivery")),
   managedSubmissionContract: Schema7.optional(Schema7.Literal("canonical_v1")),
@@ -2391,6 +2381,11 @@ var BountyPublicResponse = Schema7.Struct({
 var CancelBountyRequest = BountyCancellationFeedbackRequest.pipe(
   Schema8.extend(GetBountyRequest)
 );
+var createBountySpec = {
+  name: "create_bounty",
+  description: `Create a one-shot task bounty for humans to apply to. Dry-run preview.fundingTotal is the total funding requirement before any existing wallet balance is applied; show that estimate to the operator before posting. Live pending_deposit responses include checkout_total, the exact hosted charge after wallet balance is applied, next to deposit_url. To target an entire country, pass location with an ISO country code and isRemoteAllowed=false while omitting city and state. Standard application bounties support application detail items via applicationDetails, including text questions, optional or required acknowledgment checkboxes, one-file uploads (allowedFileTypes: image, docx, pdf, txt, video, audio), and one required live_video field whose label is the script applicants record with the in-browser camera. For appearance-based bounties where the worker appears on camera or their presence is part of the deliverable (sign holding, UGC, sponsored posts, on-camera video, promo/GTM appearances), always include one required live_video field with a label asking the applicant to explain in two sentences why they are a good fit. Set liveCaptureRequirement to photo or video when accepted workers must capture new camera evidence after acceptance; neither camera-only workflow is cryptographic liveness verification. Do not ask for passwords, OTP/2FA codes, API keys, private keys, seed phrases, government IDs, bank/card details, exact home addresses, dates of birth, or other sensitive personal information. Blank rows are ignored; uploads are capped at 3 fields, acknowledgments at 5 fields, and live video at 1 field. Direct-review photo, video, or document collection bounties use submissionMode and matching upload settings instead of applicationDetails. lifecycleMessages can define auto-message templates for acceptance, rejection, and submission review transitions. Set completionWindowHours (any whole number of hours from 1 to 720, e.g. 36) to give confirmed workers a completion deadline: overdue seats are automatically released and reopened for other applicants (auto-reassign), workers can request extensions, and you approve/deny/grant them per worker. Pass optional \`idempotencyKey\` to make this safe to retry (a replayed key returns the original result instead of duplicating). Applicants are auto-reviewed by default (autoAccept, default true): deterministic checks always run and an AI review runs only for free-text screening answers \u2014 qualified applicants are accepted, clear mismatches rejected with a reason, uncertain cases left pending for manual review; pass autoAccept: false to review and accept every application yourself; on micCheckRequired bounties, set autoAcceptMinMicScore (1-5) to make the mic score a hard cutoff: applications below it are auto-rejected, at/above proceed toward acceptance, and unscored recordings stay pending. When omitted, the platform default 3.0 applies and lower scores only stay pending for manual review. To keep previous participants out, pass excludedParticipantSourceBountyIds with up to ${MAX_EXCLUDED_PARTICIPANT_SOURCE_BOUNTIES} ids of your OWN earlier bounties: any worker who was ever accepted on one of them cannot see or apply to the new bounty, and you can change the list later with update_bounty. AI-managed, QA, and taste run bounties, and never-funded drafts are rejected as sources. BETA: pass aiManaged: true to have the platform fully manage the bounty \u2014 automated recruiting, vetting, submission review, payment release, and a final report. Fixed price USD only, 1-50 spots. Poll the bounty's managed run or subscribe to run.report_ready for the finished work.`,
+  input: CreateBountyRequest
+};
 var getBountySpec = {
   name: "get_bounty",
   description: "Get detailed information about a specific bounty, including full description, requirements, completionCriteria, evidenceTypes, evidenceCriteria, and liveCaptureRequirement. Inspect those criteria before reviewing evidence. Automated submission findings are advisory and never approve work or release payment.",
@@ -2409,7 +2404,7 @@ var getBountyDatasetSpec = {
 };
 var acceptApplicationSpec = {
   name: "accept_application",
-  description: "Accept a human's application for your bounty. REQUIRES a funded escrow for this specific application \u2014 call create_escrow_checkout first to fund it, otherwise this returns 402 Payment Required. Once funded, accepting creates a booking and locks the escrow until work is delivered. For multi-person bounties, fund and accept each applicant separately. Other applications are auto-rejected only when the bounty is fully filled. Pass optional `idempotencyKey` to make this safe to retry (a replayed key returns the original result instead of duplicating). Check the applicant's `reviewSummary` from get_bounty_applications before accepting; a null averageRating means no history, not a bad record.",
+  description: "Accept a human's application for your bounty. REQUIRES a funded escrow for this specific application \u2014 call create_escrow_checkout first to fund it, otherwise this returns 402 Payment Required. Once funded, accepting creates a booking and locks the escrow until work is delivered. For multi-person bounties, fund and accept each applicant separately. Other pending applications are kept even when every seat fills \u2014 a filled seat can still be released and refilled \u2014 and are auto-rejected only once every seat's escrow is settled (released/paid). Pass optional `idempotencyKey` to make this safe to retry (a replayed key returns the original result instead of duplicating). Check the applicant's `reviewSummary` from get_bounty_applications before accepting; a null averageRating means no history, not a bad record.",
   input: AcceptApplicationRequest
 };
 var rejectApplicationSpec = {
@@ -2429,7 +2424,7 @@ var decideExtensionSpec = {
 };
 var updateBountySpec = {
   name: "update_bounty",
-  description: "Update ordinary one-shot bounty details. You can modify the title, description, price, application-cutoff deadline, location, requiredLinks, applicationDetails, lifecycleMessages, liveCaptureRequirement, reactivate hidden inactive bounties, and more. The deadline stops new applications and direct uploads; completionWindowHours controls post-acceptance completion timing. To target an entire country, pass location with an ISO country code and isRemoteAllowed=false while omitting city and state. Live capture and required-link requirements can only be changed before applications are received. applicationDetails may be changed after applications are received; existing applications keep their original answers, while future applications use the latest fields. applicationDetails are application detail items for standard application bounties only; blank rows are ignored, uploads are capped at 3 fields, acknowledgments at 5 fields, and camera-only live_video at 1 required field. lifecycleMessages can define auto-message templates for acceptance, rejection, and submission review transitions. Admin-only ongoing bounty settings are intentionally not exposed through MCP. You can also pause/unpause a bounty (status 'paused'/'open'), close an unassigned bounty and return its unused funding (status 'closed'), increase seats via spotsAvailable, keep pending applicants on fill via keepApplicantsOnFill, and change the auto-reassign completion deadline via completionWindowHours (any whole number of hours from 1 to 720; null disables; only affects seats confirmed after the edit), and toggle automatic applicant review via autoAccept (applications received while off stay pending; toggling on affects future applications only; ignored on aiManaged bounties) or adjust its mic-quality floor via autoAcceptMinMicScore (1-5, null resets to the 3.0 default). Use cancel_bounty for cancellation and refund handling. Work completion and payment are system-managed from escrow and payout evidence; status 'completed' and 'paid' cannot be set directly.",
+  description: `Update ordinary one-shot bounty details. You can modify the title, description, price, application-cutoff deadline, location, requiredLinks, applicationDetails, lifecycleMessages, liveCaptureRequirement, reactivate hidden inactive bounties, and more. The deadline stops new applications and direct uploads; completionWindowHours controls post-acceptance completion timing. To target an entire country, pass location with an ISO country code and isRemoteAllowed=false while omitting city and state. Live capture and required-link requirements can only be changed before applications are received. applicationDetails may be changed after applications are received; existing applications keep their original answers, while future applications use the latest fields. applicationDetails are application detail items for standard application bounties only; blank rows are ignored, uploads are capped at 3 fields, acknowledgments at 5 fields, and camera-only live_video at 1 required field. lifecycleMessages can define auto-message templates for acceptance, rejection, and submission review transitions. Admin-only ongoing bounty settings are intentionally not exposed through MCP. You can also pause/unpause a bounty (status 'paused' stops new applications and outreach, including on a fully staffed bounty; status 'open' resumes it, and the resumed status is derived from its seats: open, partially_filled, or assigned when every seat is still filled), close an unassigned bounty and return its unused funding (status 'closed'), increase seats via spotsAvailable, keep pending applicants on fill via keepApplicantsOnFill, and change the auto-reassign completion deadline via completionWindowHours (any whole number of hours from 1 to 720; null disables; only affects seats confirmed after the edit), and toggle automatic applicant review via autoAccept (applications received while off stay pending; toggling on affects future applications only; ignored on aiManaged bounties) or adjust its mic-quality cutoff via autoAcceptMinMicScore (1-5 \u2014 applications scoring below your configured cutoff are auto-rejected, at/above proceed; null resets to the 3.0 default, below which applications only stay pending). The exclude-previous-participants list can be changed after posting: pass excludedParticipantSourceBountyIds with the full new list of up to ${MAX_EXCLUDED_PARTICIPANT_SOURCE_BOUNTIES} ids of your OWN earlier bounties (an empty array clears it); the same source rules as create_bounty apply, and a bounty can never exclude itself. Use cancel_bounty for cancellation and refund handling. Work completion and payment are system-managed from escrow and payout evidence; status 'completed' and 'paid' cannot be set directly.`,
   input: UpdateBountyRequest
 };
 var cancelBountySpec = {
@@ -2439,12 +2434,12 @@ var cancelBountySpec = {
 };
 var getBountyOutreachStatusSpec = {
   name: "get_bounty_outreach_status",
-  description: "See the automatic outreach the platform ran for one of your bounties: how many candidate humans were contacted (with timestamps), campaign progress toward the send target, and how many contacted humans applied or responded. Owner only. Use this to check whether a bounty is getting in front of qualified humans before deciding to boost.",
+  description: "See the automatic outreach the platform ran for one of your bounties: how many candidate humans were emailed (with timestamps), how many alerts are still queued, progress toward the delivered-alert target, and how many contacted humans applied or responded. Owner only. Use this to check whether a bounty is getting in front of qualified humans before deciding to boost.",
   input: GetBountyRequest
 };
 var boostBountyOutreachSpec = {
   name: "boost_bounty_outreach",
-  description: "Re-run relaxed candidate outreach for one of your open bounties right now, contacting additional matched humans instead of waiting for the hourly automatic sweep. Owner only. Rate-limited to once per bounty per 24 hours. Use get_bounty_outreach_status first to see current reach.",
+  description: "Raise the outreach target for one of your open bounties by another wave of matched humans (up to 300 more, capped at 1250 lifetime) and run that wave right now. Owner only. Rate-limited to once per bounty per 24 hours. Use get_bounty_outreach_status first to see current reach.",
   input: GetBountyRequest
 };
 var payEnterpriseBountySpec = {
@@ -3225,13 +3220,13 @@ var humans = {
   notFound: "Human not found"
 };
 var conversations = {
-  verificationRequired: "Direct messaging requires account eligibility. Searching humans and posting bounties are available with an API key; use create_bounty to start a pay-to-play task instead of cold messaging.",
+  verificationRequired: "Direct messaging is disabled. Use create_bounty to start paid work, then message applicants in its conversations.",
   verificationNextStep: "Use create_bounty when you are ready to hire."
 };
 var onboarding = {
   accountStatusNoKey: "No API key configured. Create one at rentahuman.ai/account/api-keys and set it as RENTAHUMAN_API_KEY.",
   accountStatusNoKeyNextStep: "Open rentahuman.ai/account/api-keys, create a key, then rerun rentahuman-mcp setup or set RENTAHUMAN_API_KEY in your MCP config.",
-  accountStatusApiKeyConfigured: "API key configured. You can search humans and post bounties; direct messaging is controlled by account eligibility and existing conversation context.",
+  accountStatusApiKeyConfigured: "API key configured. You can search humans, post bounties, and message supported task or platform-support conversations. Direct conversation creation is retired.",
   accountStatusKeyInvalid: "API key is configured but could not verify status. The key may have been revoked. Check rentahuman.ai/account/api-keys."
 };
 var identity = {
@@ -3243,7 +3238,7 @@ var identity = {
     "Use list_identities to see all available identities.",
     "Use switch_identity to change to a different identity.",
     ...hasApiKey ? [
-      "API key is configured. Write operations (bounties, conversations, messages) are enabled.",
+      "API key is configured. Bounty writes and messages in supported task or platform-support conversations are enabled.",
       "To link this crypto identity with your API key, call POST /api/keys/register-identity with your agentId, publicKey, signature, and timestamp (signed with action 'register_identity')."
     ] : [
       "WARNING: No API key configured. Create one at rentahuman.ai/account/api-keys and set RENTAHUMAN_API_KEY in your MCP config."
@@ -3320,6 +3315,8 @@ var handleCreateBounty = (args) => Effect9.gen(function* () {
     return err({ success: false, error: screeningFieldsResult.error });
   }
   const screeningFields = screeningFieldsResult.fields;
+  const excludedParticipantSourceBountyIds = params.excludedParticipantSourceBountyIds?.length ? params.excludedParticipantSourceBountyIds : void 0;
+  const exclusionFields = excludedParticipantSourceBountyIds ? { excludedParticipantSourceBountyIds } : {};
   if (params.dryRun) {
     const spotsAvailable = params.spotsAvailable || 1;
     const currency = params.currency || "USD";
@@ -3350,6 +3347,7 @@ var handleCreateBounty = (args) => Effect9.gen(function* () {
       keepApplicantsOnFill: params.keepApplicantsOnFill || false,
       completionWindowHours: params.completionWindowHours ?? null,
       requiredLinks: params.requiredLinks || [],
+      ...exclusionFields,
       applicationDetails: screeningFields,
       submissionMode: params.submissionMode || "application",
       photoSubmission: params.photoSubmission || null,
@@ -3406,6 +3404,7 @@ var handleCreateBounty = (args) => Effect9.gen(function* () {
       keepApplicantsOnFill: params.keepApplicantsOnFill || false,
       ...params.completionWindowHours !== void 0 ? { completionWindowHours: params.completionWindowHours } : {},
       requiredLinks: params.requiredLinks || [],
+      ...exclusionFields,
       applicationDetails: screeningFields,
       submissionMode: params.submissionMode || "application",
       photoSubmission: params.photoSubmission,
@@ -3429,10 +3428,15 @@ var handleCreateBounty = (args) => Effect9.gen(function* () {
       })
     );
   }
-  const { dryRun: _, ...createParams } = params;
+  const {
+    dryRun: _,
+    excludedParticipantSourceBountyIds: _rawExclusion,
+    ...createParams
+  } = params;
   const verification = yield* identity2.createVerification("create_bounty");
   const result = yield* api.post("/bounties", {
     ...createParams,
+    ...exclusionFields,
     applicationDetails: screeningFields,
     ...verification
   });
@@ -3743,7 +3747,7 @@ var acceptApplicationTool = toMcpTool(
             spotsFilled,
             spotsRemaining
           },
-          rejectedCount: isFullyFilled ? 1 : 0,
+          rejectedCount: 0,
           mode: "mock"
         })
       );
@@ -3904,7 +3908,7 @@ var payEnterpriseBountyTool = toMcpTool(
 var bountyTools = [
   {
     name: "create_bounty",
-    description: "Create a one-shot task bounty for humans to apply to. To target an entire country, pass location with an ISO country code and isRemoteAllowed=false while omitting city and state. **IMPORTANT: Always call with dryRun=true first** to preview the bounty. Dry-run `preview.fundingTotal` is the total funding requirement before any existing wallet balance is applied. Show that estimate to the operator before posting. Show the preview to the user and ask 'Here's your bounty \u2014 would you like to edit anything before posting?' Only call again with dryRun=false (or omitted) after the user confirms. **You MUST help the user define completionCriteria and evidenceTypes** \u2014 ask what 'done' looks like and what proof they need (text/data, photos, video, or links). For specialized standard-application tasks, configure applicationDetails so applicants provide application details before review. Blank rows are ignored; application upload fields are one file each and capped at 3 total; acknowledgment checkboxes are capped at 5 total and can be optional or required; and one required live_video field may contain a script applicants must record with the in-browser camera. For appearance-based bounties where the worker appears on camera or their presence is part of the deliverable (sign holding, UGC, sponsored posts, on-camera video, promo/GTM appearances), always include one required live_video field with a label asking the applicant to explain in two sentences why they are a good fit \u2014 this gates applications on a live self-recorded video so the poster sees each applicant before accepting. Do not ask for passwords, OTP/2FA codes, API keys, private keys, seed phrases, government IDs, bank/card details, exact home addresses, dates of birth, or other sensitive personal information. For direct-review collection bounties, use submissionMode='photo_upload', 'video_upload', or 'document_upload' with matching submission settings instead of applicationDetails. lifecycleMessages can define auto-message templates for acceptance, rejection, and submission review transitions. You can require applicants to provide specific links (LinkedIn, GitHub, resume, etc.) using the requiredLinks parameter. Requires RENTAHUMAN_API_KEY from the account owner. Standard accounts use available wallet balance first; if the wallet cannot cover the bounty, the response includes deposit_url and checkout_total (the hosted charge after wallet balance is applied) and the account owner must complete checkout before the bounty is visible. Supports multi-person bounties by setting spotsAvailable > 1. Set completionWindowHours (any whole number of hours from 1 to 720, e.g. 36) to give confirmed workers a completion deadline: overdue seats auto-release and reopen for other applicants; workers may request extensions you answer with decide_extension_request. Ongoing data-collection programs are managed separately and cannot be created through this tool. Pass optional `idempotencyKey` to make this safe to retry (a replayed key returns the original result instead of duplicating). Applicants are auto-reviewed by default (autoAccept, default true): deterministic checks always run and an AI review runs only for free-text screening answers \u2014 qualified applicants are accepted, clear mismatches rejected with a reason, uncertain cases left pending for manual review; pass autoAccept: false to review and accept every application yourself; on micCheckRequired bounties, tune the mic-quality floor with autoAcceptMinMicScore (1-5, default 3.0). BETA: pass aiManaged: true to have the platform fully manage the bounty \u2014 automated recruiting, vetting, submission review, payment release, and a final report. Fixed price USD only, 1-50 spots. Poll the bounty's managed run or subscribe to run.report_ready for the finished work.",
+    description: `Create a one-shot task bounty for humans to apply to. To target an entire country, pass location with an ISO country code and isRemoteAllowed=false while omitting city and state. **IMPORTANT: Always call with dryRun=true first** to preview the bounty. Dry-run \`preview.fundingTotal\` is the total funding requirement before any existing wallet balance is applied. Show that estimate to the operator before posting. Show the preview to the user and ask 'Here's your bounty \u2014 would you like to edit anything before posting?' Only call again with dryRun=false (or omitted) after the user confirms. **You MUST help the user define completionCriteria and evidenceTypes** \u2014 ask what 'done' looks like and what proof they need (text/data, photos, video, or links). For specialized standard-application tasks, configure applicationDetails so applicants provide application details before review. Blank rows are ignored; application upload fields are one file each and capped at 3 total; acknowledgment checkboxes are capped at 5 total and can be optional or required; and one required live_video field may contain a script applicants must record with the in-browser camera. For appearance-based bounties where the worker appears on camera or their presence is part of the deliverable (sign holding, UGC, sponsored posts, on-camera video, promo/GTM appearances), always include one required live_video field with a label asking the applicant to explain in two sentences why they are a good fit \u2014 this gates applications on a live self-recorded video so the poster sees each applicant before accepting. Do not ask for passwords, OTP/2FA codes, API keys, private keys, seed phrases, government IDs, bank/card details, exact home addresses, dates of birth, or other sensitive personal information. For direct-review collection bounties, use submissionMode='photo_upload', 'video_upload', or 'document_upload' with matching submission settings instead of applicationDetails. lifecycleMessages can define auto-message templates for acceptance, rejection, and submission review transitions. You can require applicants to provide specific links (LinkedIn, GitHub, resume, etc.) using the requiredLinks parameter. Requires RENTAHUMAN_API_KEY from the account owner. Standard accounts use available wallet balance first; if the wallet cannot cover the bounty, the response includes deposit_url and checkout_total (the hosted charge after wallet balance is applied) and the account owner must complete checkout before the bounty is visible. Supports multi-person bounties by setting spotsAvailable > 1. Set completionWindowHours (any whole number of hours from 1 to 720, e.g. 36) to give confirmed workers a completion deadline: overdue seats auto-release and reopen for other applicants; workers may request extensions you answer with decide_extension_request. Pass optional \`idempotencyKey\` to make this safe to retry (a replayed key returns the original result instead of duplicating). Applicants are auto-reviewed by default (autoAccept, default true): deterministic checks always run and an AI review runs only for free-text screening answers \u2014 qualified applicants are accepted, clear mismatches rejected with a reason, uncertain cases left pending for manual review; pass autoAccept: false to review and accept every application yourself; on micCheckRequired bounties, set autoAcceptMinMicScore (1-5) to make the mic score a hard cutoff: applications below it are auto-rejected, at/above proceed toward acceptance, and unscored recordings stay pending. When omitted, the platform default 3.0 applies and lower scores only stay pending for manual review. To keep previous participants out, pass excludedParticipantSourceBountyIds with up to ${MAX_EXCLUDED_PARTICIPANT_SOURCE_BOUNTIES} ids of your OWN earlier bounties: any worker who was ever accepted on one of them cannot see or apply to the new bounty, and you can change the list later with update_bounty. AI-managed, QA, and taste run bounties, and never-funded drafts are rejected as sources. BETA: pass aiManaged: true to have the platform fully manage the bounty \u2014 automated recruiting, vetting, submission review, payment release, and a final report. Fixed price USD only, 1-50 spots. Poll the bounty's managed run or subscribe to run.report_ready for the finished work.`,
     inputSchema: toInputSchema(CreateBountyRequest),
     handler: handleCreateBounty
   },
@@ -4297,64 +4301,20 @@ var ListConversationsSchema = Schema11.extend(
 // src/tools/handlers/conversations.ts
 var startConversation = {
   name: "start_conversation",
-  description: "Start a conversation with a human. Direct messaging requires account eligibility; searching humans and posting bounties are available with an API key. For first contact, prefer create_bounty so work starts from a funded task instead of cold messaging.",
+  description: "Retired: direct messaging is disabled. Create a bounty and message applicants in its conversations instead. This tool returns an error.",
   inputSchema: toInputSchema(StartConversationSchema),
-  handler: (args) => Effect11.gen(function* () {
-    const params = yield* decode(StartConversationSchema)(args);
-    const identity2 = yield* Identity;
-    const apiKey = yield* identity2.getApiKey();
-    if (!apiKey) {
-      return err({
-        success: false,
-        error: messages.conversations.verificationRequired,
-        error_code: "verification_required",
-        upgrade_url: "https://rentahuman.ai/account",
-        nextStep: messages.conversations.verificationNextStep
-      });
-    }
-    const verification = yield* identity2.createVerification(
-      `conversation:${params.humanId}`
-    );
-    const rateLimit = yield* RateLimit;
-    yield* rateLimit.check("start_conversation");
-    const config = yield* McpConfig;
-    if (config.mockMode) {
-      const human = MOCK_HUMANS.find((h) => h.id === params.humanId);
-      if (!human) {
-        return err({ success: false, error: messages.humans.notFound });
-      }
-      const conversationId = `conv_${Date.now()}`;
-      const conversation = {
-        id: conversationId,
-        humanId: params.humanId,
-        humanName: human.name,
-        agentId: verification.agentId,
-        agentName: params.agentName || "AI Agent",
-        agentType: params.agentType,
-        subject: params.subject,
-        status: "active",
-        messageCount: 1,
-        unreadByHuman: 1,
-        lastMessageAt: (/* @__PURE__ */ new Date()).toISOString(),
-        createdAt: (/* @__PURE__ */ new Date()).toISOString(),
-        agentVerification: verification.agentVerification
-      };
-      return withFollowUpHint(
-        ok({ success: true, conversation, mode: "mock" })
-      );
-    }
-    const api = yield* ApiClient;
-    const result = yield* api.post("/conversations", {
-      ...params,
-      agentId: verification.agentId,
-      agentVerification: verification.agentVerification
-    });
-    return withFollowUpHint(ok(result));
-  })
+  handler: () => Effect11.succeed(
+    err({
+      success: false,
+      error: "Direct messaging is no longer available. Post a bounty to connect with humans.",
+      error_code: "direct_messaging_disabled",
+      nextStep: "Use create_bounty to post a task. Messaging remains available for bounty, service-booking, and support conversations."
+    })
+  )
 };
 var sendMessage = {
   name: "send_message",
-  description: "Send a message in an existing conversation. Your agent identity is cryptographically verified to prevent impersonation. Pass optional `idempotencyKey` to make this safe to retry (a replayed key returns the original result instead of duplicating).",
+  description: "Send a message in a bounty, service-booking, feedback, dispute, or platform-support conversation. Legacy direct threads are read-only. Your agent identity is cryptographically verified to prevent impersonation. Pass optional `idempotencyKey` to make this safe to retry (a replayed key returns the original result instead of duplicating).",
   inputSchema: toInputSchema(SendMessageSchema),
   handler: (args) => Effect11.gen(function* () {
     const params = yield* decode(SendMessageSchema)(args);
@@ -4502,7 +4462,7 @@ var listConversations = {
 };
 var getReviews = {
   name: "get_reviews",
-  description: "Get reviews for a specific human. Use this to check a human's reputation before starting a conversation. Supports cursor-based pagination.",
+  description: "Get reviews for a specific human. Use this to check a human's reputation before hiring. Supports cursor-based pagination.",
   inputSchema: toInputSchema(GetReviewsRequest),
   handler: (args) => Effect11.gen(function* () {
     const params = yield* decode(GetReviewsRequest)(args);
@@ -5069,7 +5029,7 @@ var GetHumanizationRequest = Schema15.Struct({
 // ../contracts/src/tool-specs/humanizations.ts
 var createHumanizationSpec = {
   name: "create_humanization",
-  description: 'Hire one human to rewrite private text without generative AI. Requires a fixed worker price, turnaround, transformation goal, and stable idempotencyKey. Provide a short `subject` (PUBLIC \u2014 shown in the listing title, e.g. "essay on AI infrastructure"; never put private content in it) so workers know what kind of text they are applying to rewrite. Optionally provide applicantScreening.aiGeneratedSourceText to require a proportional pre-acceptance rewrite sample (capped at 500 words) with its own document, continuous screen recording, and Pangram analysis. A resume upload can also be required. The main source is revealed only to the accepted worker. Pangram results are advisory and never prove authorship. Do not use this tool to falsify credentials, impersonate recommendation authors, violate academic-integrity rules, or evade required AI disclosure. BETA: pass aiManaged: true to have the platform recruit and vet writers, accept one, review the delivered rewrite and AI-detection analysis, release payment for approved work, refund unused budget, and report back. Requires USD and turnaroundMinutes >= 120; poll get_humanization for progress and the final report.',
+  description: 'Hire one human to rewrite private text without generative AI. Requires a fixed worker price, turnaround, transformation goal, and stable idempotencyKey. Provide a short `subject` (PUBLIC \u2014 shown in the listing title, e.g. "essay on AI infrastructure"; never put private content in it) so workers know what kind of text they are applying to rewrite. Optionally provide applicantScreening.aiGeneratedSourceText to require a proportional pre-acceptance rewrite sample (capped at 500 words) with its own document, continuous screen recording, and Pangram analysis. A valid Pangram AI fraction below 50% automatically approves the screening; 50% or greater rejects it, while missing or invalid scores remain in manual review. A resume upload can also be required. The main source is revealed only to the accepted worker. Pangram scores are evidence, not proof of authorship, and never release payment by themselves. Do not use this tool to falsify credentials, impersonate recommendation authors, violate academic-integrity rules, or evade required AI disclosure. BETA: pass aiManaged: true to have the platform recruit and vet writers, accept one, review the delivered rewrite and AI-detection analysis with the same threshold, release payment for approved work, refund unused budget, and report back. Requires USD and turnaroundMinutes >= 120; poll get_humanization for progress and the final report.',
   input: Schema16.Struct({
     ...CreateHumanizationRequest.fields,
     idempotencyKey: Schema16.String.pipe(
@@ -5081,7 +5041,7 @@ var createHumanizationSpec = {
 };
 var getHumanizationSpec = {
   name: "get_humanization",
-  description: "Get an owned humanization bounty, including funding and assignment state, deadline attempts, submitted document and screen recording, review state, and advisory Pangram results.",
+  description: "Get an owned humanization bounty, including funding and assignment state, deadline attempts, submitted document and screen recording, applicant screening and delivery review state, and Pangram results.",
   input: GetHumanizationRequest
 };
 
@@ -5273,7 +5233,7 @@ import { Schema as Schema18 } from "effect";
 var ListSkillsRequest = Schema18.Struct({});
 
 // ../contracts/src/tool-specs/humans.ts
-var browseTasteHumansInstruction = "Use a human's id with start_conversation or rent_human, or open profileUrl to view their profile.";
+var browseTasteHumansInstruction = "Use a human's id with rent_human or create_bounty, or open profileUrl to view their profile. Direct conversation creation is retired.";
 var browseTasteHumansSpec = {
   name: "browse_taste_humans",
   description: "Browse humans with creative/taste skills (design, visual art, music, photo & video, fashion & style, writing & performance), curated for work needing human aesthetic judgment.",
@@ -5281,7 +5241,7 @@ var browseTasteHumansSpec = {
 };
 var blockHumanSpec = {
   name: "block_human",
-  description: "Block a human profile from applying to your bounties or starting direct conversations with your account. Blocked humans can never be auto-hired (auto-accepted) onto your bounties and are excluded from your bounty outreach. Blocking a human removes them from your preferred humans list. Requires RENTAHUMAN_API_KEY.",
+  description: "Block a human profile from applying to your bounties. Blocked humans can never be auto-hired (auto-accepted) onto your bounties and are excluded from your bounty outreach. Blocking a human removes them from your preferred humans list. Requires RENTAHUMAN_API_KEY.",
   input: BlockHumanRequest
 };
 var unblockHumanSpec = {
@@ -5841,7 +5801,7 @@ var DeleteIdentitySchema = Schema20.Struct({
 // src/tools/handlers/identity.ts
 var getAgentIdentity = {
   name: "get_agent_identity",
-  description: "Get your cryptographic agent identity. This returns your unique agent ID (derived from your public key) and credentials for signing messages. Your agent ID cannot be impersonated by other agents because it's cryptographically tied to your private key. Also shows which named identity is currently active. Call this first to get your verified identity before starting conversations.",
+  description: "Get your cryptographic agent identity. This returns your unique agent ID (derived from your public key) and credentials for signing messages. Your agent ID cannot be impersonated by other agents because it's cryptographically tied to your private key. Also shows which named identity is currently active. Call this first before sending messages in supported conversations.",
   inputSchema: {
     type: "object",
     properties: {}
@@ -6100,8 +6060,8 @@ var checkAccountStatus = {
         capabilities: {
           search: true,
           bounties: true,
-          conversations: "requires direct-message eligibility",
-          messages: "requires an existing conversation or eligibility"
+          conversations: "read existing and bounty-linked conversations",
+          messages: "requires a supported task or support conversation"
         },
         message: messages.onboarding.accountStatusApiKeyConfigured
       });
@@ -7992,6 +7952,7 @@ var allTools = [
 ];
 function getVisibleTools(exposeIdentityTools, identityCount) {
   return allTools.filter((tool) => {
+    if (tool.name === "start_conversation") return false;
     if (IDENTITY_TOOL_NAMES.has(tool.name) && !exposeIdentityTools && identityCount <= 1) {
       return false;
     }

package/dist/types/index.d.ts

@@ -75,6 +75,7 @@ declare const CreateBountyRequest: Schema.Struct<{
     autoAccept: Schema.optional<typeof Schema.Boolean>;
     autoAcceptMinMicScore: Schema.optional<Schema.filter<Schema.filter<typeof Schema.Number>>>;
     identityRequired: Schema.optional<typeof Schema.Boolean>;
+    excludedParticipantSourceBountyIds: Schema.optional<Schema.filter<Schema.Array$<typeof Schema.String>>>;
     micCheckRequired: Schema.optional<typeof Schema.Boolean>;
     aiManaged: Schema.optional<Schema.Literal<[true]>>;
     asOrg: Schema.optional<Schema.Literal<[true]>>;
@@ -146,6 +147,7 @@ declare const UpdateBountyRequest: Schema.Struct<{
     }>>;
     status: Schema.optional<Schema.Literal<["open", "in_review", "paused", "closed"]>>;
     identityRequired: Schema.optional<typeof Schema.Boolean>;
+    excludedParticipantSourceBountyIds: Schema.optional<Schema.filter<Schema.Array$<typeof Schema.String>>>;
     micCheckRequired: Schema.optional<typeof Schema.Boolean>;
     responseWindowHours: Schema.optional<Schema.filter<Schema.filter<typeof Schema.Number>>>;
     autoExpireGhosts: Schema.optional<typeof Schema.Boolean>;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "rentahuman-mcp",
-  "version": "3.1.0",
+  "version": "3.2.0",
   "description": "MCP server for AI agents to browse and book humans on rentahuman.ai",
   "keywords": [
     "ai",