YodMCP — Agent Operating System

YodMCP is a production-oriented Agent Operating System kernel on MCP: multi-graph memory, Tasks, Skills, A2A, attestation (software / simulated TEE), OpenTelemetry, and plan-based monetization.

This README is written so a senior engineer with only this file + the source tree can install, deploy every surface, exercise features, and verify end-to-end.

Package surfaces

Related MCP server: Agentic Control Framework (ACF)

Requirements

• Python 3.11 or 3.12 (3.10+ declared; CI covers 3.11/3.12)

• Optional: Docker for multi-service compose

Install

git clone https://github.com/ANAMIZED/YodMCP.git
cd YodMCP
python -m venv .venv && source .venv/bin/activate   # recommended
pip install -e ".[dev]"

Copy environment defaults:

cp .env.example .env

TEE honesty: nitro / sgx are provider hooks; without real TEE libraries they fall back to simulated claims. Prefer software or simulated_tee for local verify.

Quick start (local)

# 1) MCP over stdio (attach from an MCP client)
yodmcp

# 2) REST API
yodmcp-api --port 8080
# curl http://127.0.0.1:8080/health

# 3) A2A
yodmcp-a2a --port 9000
# curl http://127.0.0.1:9000/a2a/card

MCP client configuration (stdio)

Cursor — project .cursor/mcp.json or global MCP settings:

{
  "mcpServers": {
    "yodmcp": {
      "command": "yodmcp",
      "args": [],
      "env": {
        "YODMCP_MEMORY_BACKEND": "memory",
        "YODMCP_PLAN": "free",
        "YODMCP_ATTEST_MODE": "software"
      }
    }
  }
}

If yodmcp is not on PATH, use the module form:

{
  "mcpServers": {
    "yodmcp": {
      "command": "python",
      "args": ["-m", "yodmcp"],
      "cwd": "/absolute/path/to/YodMCP",
      "env": {
        "PYTHONPATH": "src",
        "YODMCP_MEMORY_BACKEND": "memory"
      }
    }
  }
}

Claude Desktop — same shape under mcpServers in claude_desktop_config.json.

Streamable HTTP MCP (remote clients that support URL transport):

yodmcp --http --host 0.0.0.0 --port 8000
# endpoint: http://127.0.0.1:8000/mcp

Docker

docker compose up --build
# API  :8080  A2A :9000  MCP HTTP :8000

Single API container:

docker build -t yodmcp .
docker run --rm -p 8080:8080 -e YODMCP_PLAN=free yodmcp

Verify end-to-end (no external accounts required)

# Unit + integration
YODMCP_MEMORY_BACKEND=memory YODMCP_ATTEST_MODE=software \
  PYTHONPATH=src pytest tests/ -v

# Durable SQLite + simulated TEE
YODMCP_MEMORY_BACKEND=sqlite YODMCP_MEMORY_DB=/tmp/yodmcp.db \
  YODMCP_ATTEST_MODE=simulated_tee \
  PYTHONPATH=src pytest tests/ -v

# Exhaustive substrate script
PYTHONPATH=src python scripts/verify_e2e.py

Expected: all tests green; script prints ALL E2E CHECKS PASSED.

API smoke (with yodmcp-api running):

curl -s http://127.0.0.1:8080/health
curl -s http://127.0.0.1:8080/api/skills
curl -s http://127.0.0.1:8080/api/billing/plans
curl -s http://127.0.0.1:8080/api/billing/status

SDK:

python examples/sdk_quickstart.py   # requires API on :8080

MCP tools (catalog)

Resources:

• skills://{name} — skill markdown body

• yodmcp://agent-card — agent card JSON

Skills

Built-in skills are always registered. Disk skills under skills/*/SKILL.md are loaded automatically (override via YODMCP_SKILLS_DIR). Repo ships:

• memory-hygiene, safe-tool-use, funding-usdc, repo-bootstrap

• Plus built-in long-horizon-planning

Monetization

• Soft quotas via tool gate; upgrade messaging on exhaustion

• Live Stripe Checkout when STRIPE_SECRET_KEY is set; otherwise checkout returns status: payment_link

Non-custodial USDC (preferred for agents)

Canonical: funding/addresses.json

Project layout

src/yodmcp/          # package (src-layout)
  core/              # server, substrate, context
  memory/            # in-memory + durable SQLite multi-graph
  tools/             # MCP tool registration + gate
  tasks/ skills/ cache/ security/ observability/ monetization/
  api/ a2a/ sdk/
skills/*/SKILL.md    # portable Agent Skills (loaded at runtime)
scripts/verify_e2e.py
tests/
docs/ARCHITECTURE.md

Known limitations (read before claiming production)

1. Plan cache uses lightweight embeddings; default similarity threshold is 0.68 (not production vector search).

2. TEE Nitro/SGX are hooks with simulated fallback unless you wire real providers.

3. Frontend src/yodmcp/frontend/dashboard.py still uses legacy core.runtime and is not the primary control plane — use yodmcp-api + MCP tools.

4. Quotas are soft (in-process meter); not a distributed billing ledger.

5. Processes do not share in-memory state; use sqlite backend + shared volume for multi-process durability.

License

Apache-2.0 · CONTRIBUTING · CHANGELOG · ARCHITECTURE · AGENTS.md