Convert gitgram/ and aitimemachine/ from embedded git repos (which showed as opaque submodule gitlinks on GitHub) into fully-tracked first-party content. AITM now merges its 7 vendored repos as first-party engines: - memory: beads (Go task DAG) + context7 (docs SDK), unified into saved_contexts - git: vendored C git (bundled binary) + gitnexus (in-process code intelligence) - agentic: a2a (@a2a-js/sdk) + pal (AITM-managed Python subprocess) Includes the full session's product build: Gitgram web/mobile/backend features, AITM website/CLI/MCP, the unified memory layer, engine build tooling (scripts/build-engines.sh, Dockerfile, CI), docs (README, ENGINES.md, ARCHITECTURE.md), and extensive test suites. Build artifacts, node_modules, external/, and gemini-cli are gitignored. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
AITM Agent Cards
This directory contains A2A agent cards — machine-readable JSON manifests that describe what an agent can do, how to reach it, and which tools it exposes.
What is an Agent Card?
An agent card follows the A2A (Agent-to-Agent) protocol specification. It acts as a self-describing capability advertisement: any orchestrator or peer agent can read the card and know exactly what the agent offers without needing out-of-band documentation.
Each card includes:
| Field | Purpose |
|---|---|
name |
Human-readable agent name |
description |
What the agent does and why it exists |
version |
Semantic version of the agent card / agent implementation |
url |
The endpoint where the agent accepts A2A tasks |
capabilities |
Protocol-level feature flags (streaming, push notifications, state history) |
skills |
High-level capabilities the agent can perform, each with an id, name, description, and tags |
mcpTools |
Individual MCP tools the agent exposes, each with a name and description |
Agent Cards in this Directory
gitgram-agent.json
The primary Gitgram AI Agent. Covers four skill domains:
- Checkpoint Management — create and restore deterministic repo snapshots via AITimeMachine
- Code Intelligence — impact analysis, dependency tracing, context retrieval (GitNexus graph)
- A2A Coordinator — delegate tasks to peer agents, poll results, discover the agent network
- Repository Analysis — serialize repos to LLM-readable text, resolve and query docs
The agent listens at http://localhost:3001/a2a in local development.
How to Use Agent Cards
Discovery
Any A2A-compatible orchestrator fetches GET /.well-known/agent.json (or the URL embedded in the card) to auto-discover capabilities before routing tasks.
Sending a Task
POST http://localhost:3001/a2a
Content-Type: application/json
{
"jsonrpc": "2.0",
"id": "task-001",
"method": "tasks/send",
"params": {
"id": "task-001",
"message": {
"role": "user",
"parts": [{ "text": "Create a checkpoint before the refactor" }]
}
}
}
Polling a Task
POST http://localhost:3001/a2a
Content-Type: application/json
{
"jsonrpc": "2.0",
"id": "poll-001",
"method": "tasks/get",
"params": { "id": "task-001" }
}
Listing Agents (via MCP tool)
Use the a2a_list_agents MCP tool to enumerate all registered agents at runtime — no manual card maintenance required.
Adding a New Agent Card
- Create
<agent-name>.jsonin this directory. - Follow the schema used by
gitgram-agent.json. - Register the agent's URL with the A2A discovery service so orchestrators can find it.
- Document the new card in this README under "Agent Cards in this Directory".
Schema Reference
The full A2A specification lives at: https://google.github.io/A2A/specification/