TMCRARequest access

Developer integrations / 01

Connect TMCRA to the agent stack you already run.

Use the stable HTTP contract directly, evaluate preview clients, or request an approved native-adapter pilot. The labels below describe release availability, not merely whether source code exists.

Request integration access GitHub

02 / INTEGRATION SURFACES

Start at the layer that matches your system.

Stable means the public contract is supported. Preview means testable but still evolving. Pilot means access is reviewed and the adapter is not a generally published package.

01CORE INTERFACE / STABLE

REST / OpenAPI

Use the public, versioned HTTP contract from any backend stack.

HTTPS · JSON · api.tmcra.com
02CLIENT / PREVIEW

Python SDK

Typed synchronous and asynchronous clients available for pilot evaluation; package details may change before general availability.

Python · pilot distribution
03CLIENT / PREVIEW

TypeScript SDK

A testable typed client for Node.js services and agent backends; package details may change before general availability.

TypeScript · pilot distribution
04COMPATIBILITY LAYER / PREVIEW

MCP Server

A local stdio bridge for approved test deployments. It is not yet presented as a generally available hosted integration.

MCP · local stdio
05NATIVE ADAPTER / PREVIEW

Codex

A downloadable preview that recalls before Codex answers, records completed turns, isolates projects, and can import retained Codex history.

SessionStart · UserPromptSubmit · StopOpen setup guide
06NATIVE ADAPTER / PILOT

OpenClaw

A validated source adapter supplied through approved pilots. No generally published package is promised yet.

before_prompt_build · agent_end
07NATIVE ADAPTER / PILOT

Hermes Agent

A validated source adapter supplied through approved pilots. No generally published package is promised yet.

prefetch · sync_turn

03 / PRODUCTION FLOW

A predictable path from event to model context.

  1. 01

    Keep credentials out of prompts

    Store the API Key in a protected local or server-side environment. Never place it in a prompt, browser bundle or project source file.

  2. 02

    Define stable memory scopes

    Separate user-global memory from project memory, then retain each conversation as a session inside its project.

  3. 03

    Ingest idempotently

    Submit conversation events with an idempotency key, then poll the asynchronous job to completion.

  4. 04

    Recall at model time

    Send the current task and receive structured evidence plus prompt_evidence.content for your existing model.

04 / DELIVERY CONTRACT

Your model stays yours.

TMCRA does not replace the model or Agent runtime. It supplies traceable evidence and a deterministic prompt context at the point where your system needs memory.

recall.response
{
  "evidence": [...],
  "prompt_evidence": {
    "content": "..."
  }
}

05 / PILOT

Tell us what you need to connect.

Submit your stack, use case and rollout timeline. The TMCRA team will review the application in the Internal control plane.

Open the application