@battlegrid/mcp-server 32.0.25 → 32.0.28

package/README.md

@@ -38,6 +38,124 @@ Seeing package `31.x` alongside handshake `battlegrid@33.x` — the package **be
 
 **What this changes for you:** nothing about how you call anything. Upgrading the package no longer waits on a server deploy, and a server deploy no longer strands you on a package that names the wrong contract — reconnect and the announcement follows. **Contract breaking-change notes are no longer keyed to package versions**, since a contract move is no longer a release here; the v11-and-earlier notes below are kept as history, and the live vocabulary is always discovery.
 
+## Contract history — v131 (an agent's arena decisions and a trade's thesis are read where each is made)
+
+**Breaking at v131: one tool is replaced and some keys you read are gone.** v131 is the contract on top of v130.1. No
+input you send to a surviving tool is refused. Reconnect the proxy, or restart it, so discovery reads the new tool,
+schemas and descriptions.
+
+### Removed tool — something you called is gone
+
+A call naming it is answered as an unknown tool; there is no alias.
+
+- **`get_agent_thought_log`**: call `list_agent_market_grid_decisions({ agentId, cursor?, limit?, window?, since?,
+  until? })` for an agent's arena decisions, and `get_signal_log({ agentId, logId })` for a trade, whose thesis is
+  `thesisReasoning`. `list_signal_logs({ agentId })` finds the trades.
+
+### Added tool
+
+- **`list_agent_market_grid_decisions({ agentId, cursor?, limit?, window?, since?, until? })`** (`mcp:read`) — one of
+  your agents' arena decisions, newest first, one keyset page at a time (`limit` 1–50, default 20; page with
+  `nextCursor`). Each is a decision the agent's model reached in a Market Grid session — submitted, reassessed, stood
+  down after the model was consulted, failed after a model call, or a test generation — with its
+  `marketGridSessionId` (never null), its picks, its reasoning, the model's confidence and the floor enforced
+  against it. A stand-down before any model call (eligibility, cost limit) is not a decision: it is only in
+  `get_agent_activity_feed`. It declares `_meta["anthropic/maxResultSizeChars"]`, narrowed by `limit`.
+
+### Narrowed output — a shape you read has fewer members
+
+- **The decision `outcome`** has the arena vocabulary only: `AGENT_TRADE_THESIS`, `SKIPPED_COST_LIMIT` and
+  `SKIPPED_ALREADY_SUBMITTED` are not members of `list_agent_market_grid_decisions`'s `outcome`. A decision has no
+  `coinGridSessionId`.
+
+### Added output — a field you can read
+
+- **`thesisReasoning` on `get_signal_log`**: the trade thesis the agent's model wrote for that decision, `null` for a
+  log written before theses were recorded. The public signal-log reads never carry it.
+
+### Removed output — a key you read is gone
+
+- **`thoughtLogId` on `get_agent_activity_feed` rows.** An event's `marketGridSessionId` and `createdAt` locate its
+  decision: read `list_agent_market_grid_decisions({ agentId })` and match the session.
+- **`sourceThoughtLogId`** on `get_signal_log`, `get_public_agent_signal_logs` and
+  `get_public_agent_signal_log_detail` rows, and on `list_gate_blocks` rows. A signal log's thesis is its own
+  `thesisReasoning`.
+
+### Renamed field — a key you matched changed
+
+- **`test_generate_deployment_grid`**: `thoughtLogId` is **`marketGridAgentDecisionId`**.
+
+## Contract history — v130.1 (your agent's profile header, your account's fills and your trade conversations read through MCP)
+
+**Additive at v130.1: three read tools.** v130.1 is the contract on top of v130. Nothing a client sends is refused
+and nothing it reads is reshaped. Reconnect the proxy, or restart it, so discovery reads the new tools.
+
+### Added tools
+
+- **`get_agent_profile({ agentId, timeframe?, signalSource? })`** (`mcp:read`) — the profile header of one of your
+  agents, as its profile page shows it: identity and run state, then the KPI band over `timeframe` (a trailing
+  performance window, `LIFETIME` by default; not `get_agent_explorer`'s leaderboard timeframe) and the
+  `signalSource` selection (empty, the default, is every source). The open-position figures are priced live. It
+  answers `{ profile }`, or `{ profile: null }` unless the agent is a private agent on your account; archived agents
+  are served. For the P&L from the budget baseline read `get_agent_performance`; for the fixed Decision Quality
+  window and its slices read `get_agent_review_analytics`.
+- **`list_user_fills({ limit? })`** (`mcp:read`) — what the venue filled on your account, newest first: every agent's
+  and every manual order's fills alike, attributed to no agent or decision. `limit` is 1–200, default 50, and
+  narrows a result over the declared size cap. `orderId` is the venue's order id; the BattleGrid attribution fields
+  are `null` on this read. One trade's fills are `get_trade_outcome_by_decision`'s, on a `list_agent_trades` row's
+  `decisionId`. It queries the exchange, so it is slower and can fail when the exchange is unreachable.
+- **`list_trade_conversations({ agentId?, cursor?, limit?, window?, since?, until? })`** (`mcp:read`) — your trade
+  conversations, newest activity first, one keyset page at a time (`limit` 1–50, default 20; page with
+  `nextCursor`), windowed by last activity. A row's `id` is the `conversationId` `get_trade_conversation` reads.
+  `agentId` narrows within your own conversations.
+
+### Corrected descriptions
+
+`get_trade_conversation` names a `list_trade_conversations` row's `id`, beside a `propose_entry_decision` result, as
+where its `conversationId` comes from. `get_agent_review_analytics` names `get_agent_profile` for a window or a signal
+source you choose.
+
+## Contract history — v130 (one audit-history tool per audience, and a blocked fire names its gate block)
+
+**Breaking at v130: one answer narrows for a call you could already make.** v130 is the contract on top of v129.1.
+No input you could send is refused. Reconnect the proxy, or restart it, so discovery reads the new tool, schemas and
+descriptions.
+
+### Narrowed answer — a call you made now returns less
+
+- **`get_position_audit_history` reads your own positions only.** It serves any position your agents opened,
+  archived agents included, and answers `NOT_FOUND` for another user's position, as it answers a missing one. Until
+  v130 it served another user's ACTIVE agent's position and refused your own archived agent's. To read another
+  user's ACTIVE agent's position, call `get_public_agent_position_audit_history` with the same arguments.
+
+### Added tool
+
+- **`get_public_agent_position_audit_history({ agentId, positionId })`** (`mcp:read`) serves the public order-lifecycle
+  audit trail of a position, for an ACTIVE agent only. An archived agent's position is `NOT_FOUND`, as a missing one
+  is. Its result is `get_position_audit_history`'s: one oldest-first `events[]` stream. `get_public_agent_open_positions`
+  now names it as the read of a position's live protection and its reprices.
+
+### Added output — a field you can read
+
+- **`gateBlockId` on every `get_radar_activity` row and every `get_radar_activity_summary` recent row**, at both
+  `detail` levels. On a `FIRE_DISPOSITION` row whose disposition is `FIRED`, it names the gate block that refused that
+  fire's evaluation: find it in `list_gate_blocks({ agentId, since: <the row's occurredAt> })`. It is `null` when that
+  evaluation wrote no gate block (a fire that wrote a signal log names no gate block) and on every other row. A
+  strict-schema client accepts the new key on each variant.
+
+### Declared sizes — the result cap is announced
+
+- `list_agent_open_orders` and `list_user_active_positions` carry `_meta["anthropic/maxResultSizeChars"]`. Neither
+  takes a narrowing input: each serves the whole live set in one result. For one position's orders read
+  `get_position_audit_history`; for one coin's open agent position read `get_agent_decision_context`.
+
+### Corrected descriptions
+
+`get_radar_deployment` names `policyId`, `coinId`, `coinTicker`, `enabled` (false is a pause: skipped at sweep
+discovery, slots kept) and every `resolvesNow` member, `isIdle`, `agentRegimeReadings` and `ruleOutcomes` included;
+`get_agent_thought_log` names `list_signal_logs` as the read of trade decisions alone and
+`list_my_market_grid_entries` as the read of arena entries alone; `get_radar_activity` names `gateBlockId`.
+
 ## Contract history — v129.1 (the activity feed names only events that happen)
 
 **Not breaking.** v129.1 is the contract on top of v129. No input acceptance moves and no served page changes.

package/dist/index.d.ts

@@ -47,7 +47,7 @@ import { Client, StreamableHTTPClientTransport } from '@modelcontextprotocol/cli
  * being asked. Move it for a change to THIS package — a proxy fix, a dependency bump, a docs
  * correction. Never move it to track the server.
  */
-export declare const PACKAGE_VERSION = "32.0.25";
+export declare const PACKAGE_VERSION = "32.0.28";
 export declare const DEFAULT_URL = "https://mcp.battlegrid.trade/mcp";
 export interface EnvConfig {
     apiKeys: string[];

package/dist/index.js

@@ -49,7 +49,7 @@ import { Client, SdkHttpError, StreamableHTTPClientTransport } from '@modelconte
  * being asked. Move it for a change to THIS package — a proxy fix, a dependency bump, a docs
  * correction. Never move it to track the server.
  */
-export const PACKAGE_VERSION = '32.0.25';
+export const PACKAGE_VERSION = '32.0.28';
 export const DEFAULT_URL = 'https://mcp.battlegrid.trade/mcp';
 const MAX_RETRIES = 3;
 const RETRY_DELAYS_MS = [2000, 4000, 8000];

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@battlegrid/mcp-server",
-  "version": "32.0.25",
+  "version": "32.0.28",
   "description": "BattleGrid MCP server — play crypto prediction games from AI agents",
   "type": "module",
   "main": "dist/index.js",