@battlegrid/mcp-server 31.2.109 → 31.2.112

package/README.md

@@ -24,10 +24,77 @@ 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 — v113 (one pagination shape: an opaque cursor)
+
+**Breaking at v113: nine reads page by cursor.** v113 is the contract on top of v112. Reconnect the proxy, or restart
+it, so discovery reads the new schemas.
+
+### Rejected input — something you send is no longer accepted
+
+- **`page` on the nine paginated reads** — `get_agent_thought_log`, `get_user_thought_log`,
+  `get_agent_activity_feed`, `get_user_activity_feed`, `list_gate_blocks`, `list_signal_logs`,
+  `list_entry_decisions`, `get_public_agent_signal_logs` and `get_public_agent_realized_trades`. Omit `cursor` for the
+  first page, then send the `nextCursor` the previous page returned. `page` is refused as an unrecognised key naming
+  it; nothing translates it.
+- **A cursor the server did not write** is refused naming `cursor`, before any read.
+
+### Reshaped output — something you read moved
+
+- **`total` → `nextCursor`** on the nine reads (`null` on the last page); the thought-log, activity-feed and
+  realized-trades reads also drop `page` and `limit`. `list_gate_blocks` keeps `summary`, which still spans every row.
+  Rows come newest first, ties broken by id, so a page boundary never skips or repeats a row.
+
+### Behaviour behind unchanged schemas
+
+- **`list_signal_logs` naming only `signalSource`** now narrows by it; it used to return every source.
+
+## Contract history — v112 (one input name per identity)
+
+**Breaking at v112: three input names retire.** v112 is the contract on top of v111.1. Reconnect the proxy, or restart
+it, so discovery reads the new schemas.
+
+### Rejected input — something you send is no longer accepted
+
+- **`symbol` on `get_regime_snapshot` and `get_regime_history`** — send `coinTicker` (default `BTC`).
+- **`ticker` on `get_coin_candles` and `get_coin_performance_history`** — send `coinTicker`.
+- **`marketGridSessionId` on `list_session_agent_positions`** — send `sessionId`.
+
+  Each retired name is refused as an unrecognised key naming it; nothing translates it. Responses keep `symbol` and
+  `marketGridSessionId` where they published them.
+- **An unknown coin on two filters.** `list_entry_decisions`' `coinTicker` and the public-profile `filter.coinTicker`
+  facet now resolve like every other coin input — `kPEPE` filters `PEPE`, any casing works — and an unknown coin is
+  refused naming `coinTicker` or `filter.coinTicker[i]`, where it used to answer an empty page.
+
+## Contract history — v111.1 (every read fits hosted Claude, and a large one says so)
+
+**One refusal is added at v111.1; nothing you send is refused that was accepted.** v111.1 is the contract on top of
+v111. Reconnect the proxy, or restart it, so discovery reads the new schemas and the new resource.
+
+### Refusal vocabulary
+
+- **`RESULT_TOO_LARGE`.** A read-only tool whose result would serialize past the server's result-character cap
+  (at most 150,000, hosted Claude's tool-result cap) is refused whole, never cut, with details
+  `{ limit: "result_chars", actual, maximum }`. `get_market_context` (fewer `modules`), and `get_agent_journal`,
+  `list_gate_blocks` and `list_market_grid_sessions` (a smaller `limit`) add `nextAct: "narrow"`; a read with no
+  input that asks for less adds none. A tool that writes is never refused this way. The code also joins the draft
+  diagnostics' `code` enum.
+
+### Behaviour behind unchanged schemas
+
+- **The server instructions are an index within 4,096 characters**, the length Claude reads: the rules every session
+  needs, then one routing line per workflow naming its entry tool and skills. The workflow cards move to the
+  `battlegrid://guide/workflows` resource, and the grid format stays in `battlegrid://reference/grid-format`.
+- **Ten large reads declare `_meta["anthropic/maxResultSizeChars"]`** equal to the server's cap, so Claude Code keeps
+  their results in the conversation rather than moving them to a file.
+- **Input schemas hold each reused or recursive shape once, under named `$defs`**, where they pointed into a sibling
+  property; they accept exactly what they accepted. No schema publishes `{ "not": {} }`, and four strategy outputs
+  stop publishing `conditionVerdicts`, a property no response carried.
+- **Descriptions name their parameters instead of listing the members of an input enum**; the enum is the list.
+
 ## Contract history — v111 (the draft tools take their arguments at the top level)
 
 **Breaking at v111: every draft-lifecycle tool takes its id, `draftVersion` and payload as top-level
-arguments.** v111 is the contract on top of v110.12. Reconnect the proxy, or restart it, so discovery
+arguments.** v111 is the contract on top of v110.13. Reconnect the proxy, or restart it, so discovery
 reads the new schemas.
 
 ### Rejected input — something you send is no longer accepted
@@ -3635,8 +3702,9 @@ you reach over MCP:
 > fails CI on a hand edit. Change the skill upstream and let the export lane bring it here.
 
 A client connected to BattleGrid's remote server directly receives its `initialize` instructions:
-the rules that span tools and one routing card per workflow, each naming the `battlegrid-*` skill
-above that holds the full procedure. This proxy does not relay those instructions; the skills
+the rules that span tools and one routing line per workflow, each naming the `battlegrid-*` skill
+above that holds the full procedure, with the workflow cards in the `battlegrid://guide/workflows`
+resource. This proxy does not relay those instructions; the skills
 installed beside it carry the same workflows.
 
 ## License

package/dist/index.d.ts

@@ -48,7 +48,7 @@ import { type Implementation, type Prompt, type Resource } from '@modelcontextpr
  * 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 = "31.2.109";
+export declare const PACKAGE_VERSION = "31.2.112";
 export declare const DEFAULT_URL = "https://mcp.battlegrid.trade/mcp";
 export interface EnvConfig {
     apiKeys: string[];

package/dist/index.js

@@ -51,7 +51,7 @@ import { ListToolsRequestSchema, CallToolRequestSchema, ListPromptsRequestSchema
  * 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 = '31.2.109';
+export const PACKAGE_VERSION = '31.2.112';
 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": "31.2.109",
+  "version": "31.2.112",
   "description": "BattleGrid MCP server — play crypto prediction games from AI agents",
   "type": "module",
   "main": "dist/index.js",