Use these docs in your agent

Drive the platform from a coding agent over MCP, pull the whole docs corpus as plain text, or copy any page as Markdown.

These docs, and the platform they describe, are built to be used by agents as well as people. A coding agent can use these docs three ways. It can drive the platform over MCP: read runs, reasoning, and verdicts, and write corrections back. It can read the whole docs corpus as plain text. Or it can copy any single page as Markdown. This page shows each option and the exact config to paste.

Drive the platform from your coding agent

The learnings MCP server exposes the Trust Loop as tools: your runs, the reasoning graph, per-rule verdicts, findings, improvement suggestions, cross-run insights, and the feedback write path. A coding agent with these tools reads what its agent did and fixes it against real graded evidence instead of guesswork.

Download the server: aisquare_learnings_mcp.py. One file, Python 3.9+, one dependency (pip install mcp), HTTP via the standard library. Every tool is a read except two deliberately safe writes: feedback dispositions and the deterministic per-run checkup. Nothing in it can delete or alter a run.

Install and register (Claude Code)

pip install mcp
curl -sO https://docs.aisquare.studio/tools/aisquare_learnings_mcp.py

export EXPLAINABILITY_GATEWAY_URL="https://<your-workspace>.aisquare.studio"
export EXPLAINABILITY_API_KEY="AIS_..."        # studio API key
export AISQUARE_STUDIO_ID="<your-studio-id>"

claude mcp add aisquare-learnings -- python3 aisquare_learnings_mcp.py

mcp.json (Cursor, Claude Desktop, Windsurf, VS Code)

{
  "mcpServers": {
    "aisquare-learnings": {
      "command": "python3",
      "args": ["aisquare_learnings_mcp.py"],
      "env": {
        "EXPLAINABILITY_GATEWAY_URL": "https://<your-workspace>.aisquare.studio",
        "EXPLAINABILITY_API_KEY": "AIS_...",
        "AISQUARE_STUDIO_ID": "<your-studio-id>"
      }
    }
  }
}

Reads need a studio key

The reads behind these tools are studio-scoped, so set a studio API key. A workspace ingest key sends traces but answers every read with 403 Studio ID mismatch. See Authentication.

The twelve tools, mapped to the Trust Loop:

Tool	What it returns	Loop step
`list_runs`	Recent runs, filterable by errors, policies, agent, date	Understand
`get_run_graph`	The run as nodes and edges with latency, tokens, cost	Understand
`get_reasoning`	Claims, inference chain, assumptions (RML)	Understand
`get_verdicts`	Per-rule pass and fail gates with the judge's reasoning	Prevent
`get_findings`	Machine-readable gaps with fingerprints	Fix
`send_feedback`	Dispositions a finding; trains the grader	Fix
`get_optimizations`	Heuristic improvement suggestions for one run	Fix
`studio_insights`	Cross-run patterns over the last N runs	Remember
`run_checkup`	Deterministic per-run health vs the agent's own history	Fix (coming soon)
`learnings_context`	Distilled lessons ready to inject into a prompt	Remember (coming soon)
`list_lessons`	The lesson list: candidate, active, promoted	Remember (coming soon)
`get_trends`	Cost, latency, and reliability trends with run ids	Remember (coming soon)

The four tools marked coming soon are already in the server. Until the lessons rollout reaches your gateway, they answer with a clear "not deployed on this gateway version yet". When the rollout lands they start answering, with no client-side change.

Feed the whole corpus to a model

Two routes expose these docs as plain Markdown that any model can read in one request. They follow the llms.txt convention.

/llms.txt

A compact index: the site summary plus a linked map of every page. Best when a model should pick what to read.

/llms-full.txt

The full text of every page concatenated into one file. Best for pasting the entire corpus into a context window.

Use /llms.txt when you want a small map and let the model follow links. Use /llms-full.txt when you want everything in one payload, for example to ground a one-off question or seed a retrieval index.

Copy a single page

Every page in these docs carries its own AI affordances in the header, next to the title:

Copy Markdown copies the page as clean Markdown, ready to paste into a chat or a prompt.
Open in ChatGPT, Claude, or Cursor opens that assistant with the page already loaded as context.

You can also fetch the raw Markdown of any page directly. Append .md to its URL:

https://docs.aisquare.studio/docs/getting-started/quickstart.md

The same URL without the suffix returns Markdown too when the request sends an Accept: text/markdown header, so a crawler or agent gets Markdown while a browser gets the rendered page.

Query the docs over MCP

Not active yet

A hosted docs-search MCP endpoint (GitMCP over this repository) requires the repository to be public, and it is currently private, so that endpoint does not answer today. Until it is live, point your agent at /llms.txt and the per-page Markdown above, or use the platform MCP server at the top of this page. This note will be replaced with the endpoint and config when it activates.

Questions

Building an agent on AISquare and need something these surfaces do not cover? Reach us at bots@aisquare.studio.