SuperAPI AI Directory

omega-brain-mcp

AI execution pipeline with NAEF logic constraints, VERITAS build gates, semantic vault integration and Tri-Node Cortex validations to curb hallucination.

Open website

About this tool

Omega Brain MCP is a standalone MCP server that provides task-scoped memory retrieval, explicit claim assessment, and local audit records for MCP clients. Its 2.3 standalone version replaces unsafe parameter steering and uncalibrated approval scoring with advisory alignment. Retrieval is based on indexed lexical candidates with stable features and does not claim dense semantic recall capability.

Main capabilities and uses:

  • Task-scoped memory retrieval: perform scoped queries through omega_rag_query to retrieve task-relevant memories.
  • Explicit claim assessment: explicitly assess claims.
  • Local audit records: record local audits.
  • 10-gate VERITAS build pipeline.
  • Encrypted audit ledger.
  • Cortex approval gate.
  • Cross-session memory.

Usage:

  1. Install: python -m pip install .
  2. Run: omega-brain-mcp or python omega_brain_mcp_standalone.py
  3. Add this command to the MCP client's persistent user configuration.
  4. Data is stored by default in ~/.omega-brain; for a portable desktop setup, use absolute paths.
  5. First call omega_preload_context and pass task and task_id, then use the same ID in omega_rag_query and omega_ingest.
  6. omega_integrity_status, omega_brain_report, and omega_ecosystem_status show current health status and historical verification limitations.
  7. omega_write_handoff requires conversation_id to equal the task ID.
  8. The provided Python client performs the MCP initialization handshake, limits requests, drains diagnostics, and raises server errors. For example: omega_call("omega_rag_query", query="deployment decision", task_id="example-task") for a scoped lookup.

Ecosystem and testing:

  • The companion Stenographer is used to save conversations; SSWP runs software checks. Their shared state directory and task ID must be consistent.
  • SSWP can enforce verification of Brain's operator policy receipt at its witness entry point.
  • Test commands: python -m unittest discover -s tests -p "test_contracts.py" and python -m pytest tests/test_build_gates.py.
  • The integration test suite is located in integration-tests/, using OMEGA_SUITE_ROOT to point to the sibling omega-brain, omega-stenographer, sswp checkout directories and the shared .venv.

Limitations and notes:

  • The current operational contract is in the repository documentation, for migration, permissions, evidence semantics, and recovery. Historical manuals and examples describe the old contract; the current documentation and current MCP schema take precedence.
  • Legacy network mode is not within the test guarantees of the standalone deployment.
  • Retrieval does not provide dense semantic recall.
  • Running requires two files and one dependency.
  • The documentation does not provide specific details on charges, prices, platforms, effects, or usage steps, so none are provided for now.