callx402 1.0.3 → 1.0.4

package/README.md

@@ -328,13 +328,79 @@ redemption is not currently enabled. Stripe credits use a separate redemption pa
 
 Live fee schedule: `GET https://payload-rail.fly.dev/v1/callx402/actions`
 (responds `"model":"paid on-demand per action; no subscription required"`).
-The x402-path price is the Stripe-path fee divided by 20 — for example
-`resolve` is $5.00 via Stripe, $0.25 via the x402 path. Never blind-retry a
-payment to reach a paid action: get the quote first.
+There are two prices per action, and they are not interchangeable: the
+**card price** is the Stripe schedule fee; the **USDC price** on the x402
+path is exactly that fee divided by 20 (the free quote endpoint is the
+source of truth for the USDC column). Both, side by side:
+
+| Action | Card (Stripe) | USDC (x402 path) |
+|---|---|---|
+| `diagnose` | $2.00 | $0.10 |
+| `explain` | $1.00 | $0.05 |
+| `evidence` | $2.00 | $0.10 |
+| `recover` | $10.00 | $0.50 |
+| `resolve` | $5.00 | $0.25 |
+| `preflight` | $3.00 | $0.15 |
+| `settlement_interpretation` | $3.00 | $0.15 |
+| `failure_classification` | $2.00 | $0.10 |
+| `duplicate_payment_risk` | $5.00 | $0.25 |
+
+Rail `evidence` and `explain` read back a prior paid invocation (the
+request carries that invocation's `invocation_id`), so they cannot be used
+cold; the free local forms read your local operation ledger instead.
+
+**Not sold on the rail:** `safe_retry`, `rescue`, `monitor`, and `execute`
+are local-only. They run in the self-hosted runtime (v2.0.0 tree), are
+absent from the rail catalog above, and the rail refuses them with
+`422 HOSTED_ACTION_UNAVAILABLE` — there is no Stripe or x402 purchase path
+for them. For the "is this retry safe?" question, the free local `recover`
+verdict is the available read-only answer.
+
+Never blind-retry a payment to reach a paid action: get the quote first.
 
 Problem to action map (which paid action answers which failure, with the free
 CLI/MCP path for each): [`docs/problem-map.md`](docs/problem-map.md).
 
+## Public RevRule evaluation (0.01 USDC)
+
+The rail also exposes one public RevRule evaluation endpoint. It is a
+RevRule demo, not a callx402 diagnostic action:
+`POST https://payload-rail.fly.dev/v1/x402/public/evaluate` evaluates one
+economic event against the fixed, transparent public demo graph (platform
+5% of gross; creator 70% of the remainder; referrer 25% of the remainder
+when a referrer is named in the event) for 0.01 USDC on Base.
+
+The full, correct request body is:
+
+```json
+{
+  "event": {
+    "eventId": "my-event-001",
+    "type": "SALE_COMPLETED",
+    "rail": "stripe",
+    "amountMicros": 100000000,
+    "currency": "USD",
+    "occurredAt": "2026-10-13T00:00:00.000Z"
+  },
+  "txHash": "0x<your settled 0.01 USDC payment on Base>"
+}
+```
+
+- `event.eventId` (any unique non-empty string) and `event.rail` are
+  **required**, even though the 402's advertised bazaar schema omits both.
+  Valid `rail` values: `stripe`, `ach`, `x402`, `manual`, `stablecoin`.
+- `event.type` must be one of: `SALE_COMPLETED`, `API_PAYMENT`,
+  `LICENSE_PAID`, `SUBSCRIPTION_PAYMENT`, `REFERRAL_CONVERTED`,
+  `ROYALTY_RECEIVED`, `CAMPAIGN_MILESTONE`, `MARKETPLACE_TRANSACTION`,
+  `TOUCH`.
+- The event is validated **before** payment: a malformed body gets
+  `400 INVALID_EVENT`, not a 402 (verified live: the body above, minus
+  `txHash`, returns the 402). The endpoint fills in the rest itself —
+  `graphId` is fixed to the public demo graph, and `currency` defaults to
+  `"USD"` and `occurredAt` to now when omitted.
+- Omit `txHash` to receive the x402 v2 `402` challenge; pay 0.01 USDC,
+  then retry the same body with the hash. Each txHash is single-use.
+
 ## Integrations
 
 Working entry points for agent frameworks, automation, MCP clients, and x402

package/bin/callx402.js

@@ -222,9 +222,11 @@ never executes, never repays. Requires PAYLOAD_VEYLINE_RECOVERY=1.
 
 This is the read-only answer to "is a retry safe?" — a SAFE_RETRY verdict
 here is advisory; performing the safe action stays with your own adapters.
-For the paid one-off verdict on a specific retry plan, use the 'safe_retry'
-rail action (quote first, then pay; see docs/problem-map.md). The MCP tool
-x402_recover covers the same free diagnostic.
+There is NO paid rail form of 'safe_retry': the hosted rail refuses it as
+local-only (it is absent from the rail catalog and cannot be bought via
+Stripe or x402). This free recover verdict is the read-only answer
+available without the self-hosted runtime tree (see docs/problem-map.md).
+The MCP tool x402_recover covers the same free diagnostic.
 
 Exit codes: 0 ok · 1 failure · 2 usage (missing identity/evidence) ·
 3 disabled/unreachable · 4 budget refused ·
@@ -256,14 +258,16 @@ Usage:
   callx402 "natural language intent"        intent mode (plans, never executes blindly)
   callx402 --help | --version
 
-Commands (hosted by default — zero setup):
+Commands (rail actions hosted by default — zero setup; local-only
+commands marked, they need the v2.0.0 runtime tree and are NOT sold on
+the hosted rail):
   diagnose    run the x402 doctor over supplied evidence        [--target <url|json>] [--evidence <json|@file>]
-  rescue      incident triage (auth-gated; read-only without auth) --incident <id>
+  rescue      incident triage, local only (auth-gated; read-only without auth) --incident <id>
   route       select a tool/route for a goal                   --goal <text>
   resolve     resolve settlement state from evidence           --evidence <json|@file>
   doctor      alias of diagnose
-  execute     run the intent pipeline                          --intent <text> [--max-budget N] [--dry-run] [--idempotency-key K]
-  monitor     sentinel snapshot [--once] or stream [--watch]
+  execute     run the intent pipeline, local only              --intent <text> [--max-budget N] [--dry-run] [--idempotency-key K]
+  monitor     sentinel snapshot, local only [--once] or stream [--watch]
   status      per-subsystem reachable/enabled/version report
   preflight   run MCP preflight checks
   inspect     capability-graph lookup/stats                    [--query <text>]

package/core/rail.js

@@ -156,7 +156,7 @@ async function invokeRail(command, args, opts = {}) {
     if (!interactive) {
       return {
         ok: false,
-        error: `Payment of $${price} required. Re-run with --yes to approve non-interactively, or run interactively.`,
+        error: `Payment of $${price} required. Re-run with --approve to approve non-interactively, or run interactively.`,
         quote: { action, price_usd: price, quote_id: quoteId, quote_inputs: quote.quote_inputs },
       };
     }

package/core/subsystems.js

@@ -52,6 +52,14 @@ const NOTES = {
   recovery: 'File-backed idempotency store used for duplicate-idempotencyKey dedupe.',
 };
 
+/** First line only: Node require errors append a multi-line "Require stack:"
+ * dump that helps nobody in a status report or an MCP tool result. The short
+ * form still names the missing module and stays actionable. */
+function briefError(err) {
+  const msg = err && err.message ? String(err.message) : String(err || 'unknown error');
+  return msg.split('\n')[0].trim();
+}
+
 function defaultV2Root() {
   return path.resolve(__dirname, '../../x402-paid-api-starter-kit/v2.0.0');
 }
@@ -98,7 +106,7 @@ function getSubsystem(name, { config = {}, deps = {} } = {}) {
   try {
     mod = requireFiles(v2Root, MODULE_FILES[name]);
   } catch (err) {
-    return { name, reachable: false, module: null, enabled: false, flag: null, version: null, note: NOTES[name] || null, subsystemStatus: 'unreachable', error: `require failed: ${err.message}` };
+    return { name, reachable: false, module: null, enabled: false, flag: null, version: null, note: NOTES[name] || null, subsystemStatus: 'unreachable', error: `require failed: ${briefError(err)}` };
   }
 
   let flag = null;
@@ -107,7 +115,7 @@ function getSubsystem(name, { config = {}, deps = {} } = {}) {
     flag = typeof mod.FLAG === 'string' ? mod.FLAG : null;
     enabled = typeof mod.enabled === 'function' ? mod.enabled() : false;
   } catch (err) {
-    return { name, reachable: true, module: mod, enabled: false, flag, version: null, note: NOTES[name] || null, subsystemStatus: 'disabled', error: `enabled() check failed: ${err.message}` };
+    return { name, reachable: true, module: mod, enabled: false, flag, version: null, note: NOTES[name] || null, subsystemStatus: 'disabled', error: `enabled() check failed: ${briefError(err)}` };
   }
 
   const version = mod.version || mod.RESCUE_VERSION || mod.EFFECT_PROOF_VERSION || null;

package/docs/problem-map.md

@@ -25,7 +25,13 @@ v2.0.0 runtime tree and `CALLX402_LOCAL=1`). The paid form runs on the
 production rail: quote first, payment verified exactly once, then the
 diagnostic executes server-side and the result is returned — zero local setup.
 The "Free path" column names the free equivalent; "rail only" means the action
-exists only as a paid rail action.
+exists only as a paid rail action. "Local only" means the opposite: the action
+exists ONLY locally (self-hosted runtime tree). The hosted rail refuses it
+with `422 HOSTED_ACTION_UNAVAILABLE`, it is absent from the rail catalog
+(`GET /v1/callx402/actions` lists nine actions), and there is no Stripe or
+x402 purchase path for it. Four actions are local-only: `safe_retry`,
+`rescue`, `monitor`, `execute`. Their fee cells below show the schedule
+figure for reference only — it cannot be paid on the rail.
 
 **Fees.** The table lists the Stripe-path fee per action. The x402 path
 quoted by the free quote endpoint is exactly that fee divided by 20
@@ -40,16 +46,16 @@ deterministic, not an estimate.
 | 1 | Settlement uncertainty — can't tell from current info whether the payment settled | `resolve` | $5.00 | `callx402 resolve --evidence ...` / `x402_resolve` | Unknown stays unknown; never auto-retry, never repay |
 | 2 | Lost result after payment — paid, but the result never arrived | `recover` | $10.00 | `callx402 recover <op> --identity ...` / `x402_recover` | Read-only: nothing charged, nothing executed; returns the safe recovery decision or the proven prior result, never a blind re-execution |
 | 3 | Retry uncertainty — something failed; need what/why before deciding if retry is safe | `diagnose` | $2.00 | `callx402 diagnose --evidence ...` / `x402_diagnose` | Unknown failure class returns UNKNOWN, never a guess |
-| 4 | Retry uncertainty, narrowed — is THIS specific retry safe to sign? | `safe_retry` | $5.00 | rail only (start free with `recover`: a SAFE_RETRY verdict there is the read-only answer) | A retry that would double-pay is refused |
+| 4 | Retry uncertainty, narrowed — is THIS specific retry safe to sign? | `safe_retry` | $5.00 (local only — not purchasable on the rail) | local only (self-hosted runtime; no CLI command in this package and no rail purchase path — start free with `recover`: a SAFE_RETRY verdict there is the read-only answer) | A retry that would double-pay is refused |
 | 5 | Duplicate-payment risk — a retry might sign a second authorization for the same intent | `duplicate_payment_risk` | $5.00 | rail only (start free: `resolve` the settlement, then `recover` for the retry verdict) | Ambiguous authorization state blocks the retry |
 | 6 | Need proof — ground truth before acting on any incident | `evidence` | $2.00 | `callx402 evidence <op>` / `x402_evidence` | First step in any incident; returns what is recorded, nothing invented |
 | 7 | Need explanation — plain-language state assessment before retry/refund/escalate | `explain` | $1.00 | `callx402 explain <op>` / `x402_explain` | INCOMPLETE means do not retry, do not repay |
 | 8 | Pre-commit check — intent, payment readiness, risk before an economic operation commits | `preflight` | $3.00 | `callx402 preflight` (local checks) | Blocked stays blocked until the reasons clear |
 | 9 | Failure triage — classify the failure so the right playbook runs | `failure_classification` | $2.00 | rail only (start free with `diagnose`) | Unclassifiable failures escalate to human review |
 | 10 | Settlement-record interpretation — read a raw settlement record, state what it means | `settlement_interpretation` | $3.00 | rail only (start free with `resolve --evidence ...`) | Ambiguous records are reported ambiguous |
-| 11 | Rescue a failed transaction — triage and recover the incident end to end | `rescue` | $15.00 | `callx402 rescue --incident <id>` (triage and quote only, never executes) | Never fabricates success; unreachable subsystems fail closed |
-| 12 | Watch an operation/workload for state changes, failures, anomalies | `monitor` | $5.00 | `callx402 monitor --once` (local sentinel snapshot) | An unwatched gap is reported as a gap, not as healthy |
-| 13 | Run an economic operation through the protected path (exactly-once) | `execute` | $10.00 | `callx402 execute --intent ... --dry-run` (plans only) | Over-budget intents refused with zero side effects |
+| 11 | Rescue a failed transaction — triage and recover the incident end to end | `rescue` | $15.00 (local only — not purchasable on the rail) | `callx402 rescue --incident <id>` (triage and quote only, never executes) | Never fabricates success; unreachable subsystems fail closed |
+| 12 | Watch an operation/workload for state changes, failures, anomalies | `monitor` | $5.00 (local only — not purchasable on the rail) | `callx402 monitor --once` (local sentinel snapshot) | An unwatched gap is reported as a gap, not as healthy |
+| 13 | Run an economic operation through the protected path (exactly-once) | `execute` | $10.00 (local only — not purchasable on the rail) | `callx402 execute --intent ... --dry-run` (plans only) | Over-budget intents refused with zero side effects |
 
 ## Where each surface consumes this map
 
@@ -70,9 +76,14 @@ deterministic, not an estimate.
 ## Notes
 
 - Action names are canonical in `snake_case`. The rail also accepts the
-  hyphenated aliases shown in the OpenAPI enum (`safe-retry`,
-  `duplicate-payment-risk`, `settlement-interpretation`,
-  `failure-classification`); both forms resolve identically (verified live).
+  hyphenated aliases shown in the OpenAPI enum (`duplicate-payment-risk`,
+  `settlement-interpretation`, `failure-classification`); both forms resolve
+  identically (verified live). `safe-retry` is not a rail action in either
+  spelling — it is local-only (see the two-tiers note above).
+- Rail `evidence` / `explain` read back a PRIOR paid invocation: the request
+  must carry that invocation's `invocation_id`, and the rail returns
+  `400 MISSING_INPUT` without one, so they cannot be used cold. The free
+  local forms in the map read your local operation ledger instead.
 - The x402-path quote is the valuation-layer quote, not the raw schedule;
   the fixed schedule above is the Stripe-path floor.
 - Local CLI / MCP-stdio diagnostics are free and read-only — no payment, no

package/integrations/core.py

@@ -5,7 +5,7 @@ Plain functions (no framework imports) so LangChain, CrewAI, AutoGen, or any
 agent framework can wrap them directly. stdlib only.
 
 The rail's public endpoints:
-  GET  /v1/callx402/actions            free — live 13-action fee schedule
+  GET  /v1/callx402/actions            free — live fee schedule (9 hosted actions)
   GET  /v1/callx402/quote              free — deterministic pre-execution quote
   POST /v1/callx402/actions/{action}   paid — unauthenticated -> x402 v2 402
                                        (USDC on Base); credit_id / txHash /
@@ -30,6 +30,13 @@ RAIL = os.environ.get("CALLX402_RAIL", "https://payload-rail.fly.dev").rstrip("/
 
 # Canonical action names (snake_case), verified live 2026-10-07.
 # Hyphenated aliases are also accepted by the rail; these are canonical.
+# AVAILABILITY (verified live 2026-10-13): only 9 of these 13 are hosted and
+# purchasable on the rail — diagnose, explain, evidence, recover, resolve,
+# preflight, settlement_interpretation, failure_classification,
+# duplicate_payment_risk. The other four (safe_retry, rescue, monitor,
+# execute) are LOCAL-ONLY: the rail refuses them with
+# 422 HOSTED_ACTION_UNAVAILABLE, they are absent from the rail catalog, and
+# they cannot be bought via Stripe or x402.
 ACTIONS = (
     "diagnose", "explain", "evidence", "recover", "resolve", "preflight",
     "monitor", "execute", "rescue", "settlement_interpretation", "safe_retry",
@@ -303,7 +310,9 @@ FREE_LOCAL_ALTERNATIVE = (
     "Local CLI / MCP-stdio diagnostics are free and read-only "
     "(git clone https://github.com/Payloadhq/callx402; "
     "node mcp/index.js as an MCP server). The paid rail actions above are "
-    "for hosted on-demand invocation with payment."
+    "for hosted on-demand invocation with payment. Note: safe_retry, "
+    "rescue, monitor, and execute are local-only and are NOT purchasable "
+    "on the hosted rail."
 )
 
 

package/integrations/langchain/callx402_tools.py

@@ -125,8 +125,11 @@ def callx402_explain(operation_id: str) -> str:
 def callx402_safe_retry(retry_plan: str) -> str:
     """Go/no-go verdict on ONE specific retry plan before anything is signed.
     Use when a retry is being considered and must be proven safe first.
-    Paid on-demand ($5.00 via Stripe). A retry that would double-pay is
-    refused."""
+    AVAILABILITY: safe_retry is LOCAL-ONLY — the hosted rail refuses it
+    (422 HOSTED_ACTION_UNAVAILABLE) and it cannot be bought via Stripe or
+    x402, so this tool currently surfaces that refusal. For the
+    retry-safety question today, use callx402_recover's read-only verdict.
+    A retry that would double-pay is refused."""
     return _render(invoke("safe_retry", {"retry_plan": retry_plan}, **_auth()))
 
 

package/mcp/index.js

@@ -18,13 +18,21 @@
  * Only read-only / diagnostic actions are exposed. Actions with side effects
  * (execute, rescue mutations) are intentionally NOT tools here.
  *
+ * Honest surface: ONLY x402_status works on a bare install of this package.
+ * The other five tools dispatch into the v2.0.0 subsystem tree (the x402
+ * Paid API Starter Kit — a separate product that is NOT bundled here) and
+ * return `subsystem_unreachable` until CALLX402_V2_ROOT points at that tree
+ * and the matching feature flag is enabled. These tools never call the
+ * hosted rail; paid hosted equivalents live at
+ * https://payloadhq.github.io/agents.json.
+ *
  * Usage (Claude Desktop / MCP client config):
  *   { "command": "node", "args": ["/path/to/callx402/mcp/index.js"] }
  * Env: CALLX402_CONFIG, CALLX402_V2_ROOT, CALLX402_TIMEOUT_MS (default 30000).
  */
 const readline = require('node:readline');
 
-const VERSION = '1.0.2';
+const VERSION = '1.0.4';
 const PROTOCOL_VERSION = '2024-11-05';
 const SERVER_NAME = 'callx402';
 const TIMEOUT_MS = Number(process.env.CALLX402_TIMEOUT_MS || 30000);
@@ -34,7 +42,29 @@ const core = require(process.env.CALLX402_CORE_PATH ||
 
 /* ------------------------------------------------------------------ */
 /* Tool definitions: name, description, input schema, core action.     */
+/*                                                                     */
+/* Requirement trailers, stated in every description because the       */
+/* description is what an MCP client actually shows the user:          */
+/*   TREE_REQUIRED — dispatches into the v2.0.0 subsystem tree, which  */
+/*     is NOT part of this package.                                    */
+/*   LEDGER_REQUIRED — reads the local Veyline operation ledger via    */
+/*     modules that are likewise not bundled here.                     */
 /* ------------------------------------------------------------------ */
+const TREE_REQUIRED =
+  ' Requires the v2.0.0 subsystem tree (x402 Paid API Starter Kit, a separate ' +
+  'product that is NOT included in this package): set CALLX402_V2_ROOT to the ' +
+  'tree and enable the subsystem flag, or this tool returns ' +
+  'subsystem_unreachable and does nothing. With the tree configured it is free ' +
+  'and read-only: no payment, no auth. A paid hosted equivalent runs on the ' +
+  'Payload rail — see https://payloadhq.github.io/agents.json.';
+const ledgerRequired = (flag) =>
+  ' Requires the local Veyline ledger modules (NOT included in this package; ' +
+  `enable with ${flag}=1 where the ledger lives) and reads ` +
+  'operations recorded on THIS machine — it cannot read back hosted rail ' +
+  'invocations. Until configured it returns subsystem_unreachable and does ' +
+  'nothing. When configured it is free and read-only: no payment, no auth.';
+const LEDGER_REQUIRED = ledgerRequired('PAYLOAD_VEYLINE_LEDGER');
+
 const TOOLS = [
   {
     name: 'x402_diagnose',
@@ -43,8 +73,8 @@ const TOOLS = [
       'Run the x402 doctor over supplied evidence. Use when something went ' +
       'wrong with an x402 payment or MCP/x402 interaction and you need a ' +
       'diagnosis. Input: evidence as a JSON string or object describing the ' +
-      'failure (error strings, responses, tx hashes). Read-only. ' +
-      ' Free read-only diagnostic surface: no payment, no auth, never Veyline-gated. For the paid on-demand callx402 action path, see https://payloadhq.github.io/agents.json.', 
+      'failure (error strings, responses, tx hashes). Read-only.' +
+      TREE_REQUIRED,
     inputSchema: {
       type: 'object',
       properties: {
@@ -66,8 +96,8 @@ const TOOLS = [
     description:
       'Show recorded ledger evidence for an operation: event trail, latest ' +
       'per-plane states (payment / execution / delivery), protocols involved. ' +
-      'Use FIRST in any incident — establishes ground truth before acting. Read-only. ' +
-      ' Free read-only diagnostic surface: no payment, no auth, never Veyline-gated. For the paid on-demand callx402 action path, see https://payloadhq.github.io/agents.json.', 
+      'Use FIRST in any incident — establishes ground truth before acting. Read-only.' +
+      LEDGER_REQUIRED,
     inputSchema: {
       type: 'object',
       properties: {
@@ -84,8 +114,8 @@ const TOOLS = [
       'Explain an operation\'s state in plain language from recorded evidence. ' +
       'Returns an assessment: NO_BASIS, INCOMPLETE (do not retry, do not repay), ' +
       'KNOWN_SAFE, RECOVERY_CANDIDATE, or PARTIAL. Use to understand an incident ' +
-      'before deciding whether to retry, refund, or escalate. Read-only. ' +
-      ' Free read-only diagnostic surface: no payment, no auth, never Veyline-gated. For the paid on-demand callx402 action path, see https://payloadhq.github.io/agents.json.', 
+      'before deciding whether to retry, refund, or escalate. Read-only.' +
+      LEDGER_REQUIRED,
     inputSchema: {
       type: 'object',
       properties: {
@@ -102,8 +132,8 @@ const TOOLS = [
       'Report the safe recovery decision for an operation whose result was lost ' +
       '(paid but no result). READ-ONLY: nothing is charged, nothing is executed. ' +
       'Returns RECOVERABLE/SAFE_RETRY -> KNOWN_SAFE, or HUMAN_REVIEW. Use for ' +
-      'the exact situation where blind retry would double-pay or double-execute. ' +
-      ' Free read-only diagnostic surface: no payment, no auth, never Veyline-gated. For the paid on-demand callx402 action path, see https://payloadhq.github.io/agents.json.',
+      'the exact situation where blind retry would double-pay or double-execute.' +
+      ledgerRequired('PAYLOAD_VEYLINE_RECOVERY'),
     inputSchema: {
       type: 'object',
       properties: {
@@ -119,8 +149,8 @@ const TOOLS = [
     action: 'resolve',
     description:
       'Resolve settlement state from evidence (e.g. settlement_pending: was the ' +
-      'broadcast transaction confirmed on chain?). Use instead of blind retry. Read-only. ' +
-      ' Free read-only diagnostic surface: no payment, no auth, never Veyline-gated. For the paid on-demand callx402 action path, see https://payloadhq.github.io/agents.json.', 
+      'broadcast transaction confirmed on chain?). Use instead of blind retry. Read-only.' +
+      TREE_REQUIRED,
     inputSchema: {
       type: 'object',
       properties: {
@@ -137,8 +167,11 @@ const TOOLS = [
     action: 'status',
     description:
       'Check callx402 subsystem reachability (which diagnostic subsystems are ' +
-      'available). Use to verify the tool is operational before diagnosing. Read-only. ' +
-      ' Free read-only diagnostic surface: no payment, no auth, never Veyline-gated. For the paid on-demand callx402 action path, see https://payloadhq.github.io/agents.json.', 
+      'available). Use to verify the tool is operational before diagnosing. ' +
+      'Read-only. This is the one tool that works with zero setup in any ' +
+      'install — no key, no tree, no payment. On a bare npm install it ' +
+      'honestly reports 0/15 subsystems reachable: the other five tools need ' +
+      'the v2.0.0 subsystem tree (CALLX402_V2_ROOT) before they can answer.',
     inputSchema: { type: 'object', properties: {} },
   },
 ];

package/package.json

@@ -27,6 +27,6 @@
     "smoke": "node bin/callx402.js status --json",
     "test": "node --test test/*.test.js"
   },
-  "version": "1.0.3",
+  "version": "1.0.4",
   "mcpName": "io.github.payload-tools/callx402"
 }

package/server.json

@@ -10,7 +10,7 @@
       "transport": {
         "type": "stdio"
       },
-      "version": "1.0.2"
+      "version": "1.0.3"
     }
   ],
   "repository": {
@@ -18,6 +18,6 @@
     "url": "https://github.com/Payloadhq/callx402"
   },
   "title": "callx402 MCP server",
-  "version": "1.0.2",
+  "version": "1.0.3",
   "websiteUrl": "https://github.com/Payloadhq/callx402"
 }