Learn

Documentation

Everything you need to build with Ejentum. From quickstart guides to advanced patterns.

Changelog

API changes and updates. For the full development history, see our GitHub.

v1.2. May 2026

Added: hosted MCP endpoint at api.ejentum.com/mcp.

  • The Logic API now hosts a first-party MCP endpoint at https://api.ejentum.com/mcp (Streamable-HTTP transport, Bearer auth via EJENTUM_API_KEY). Same four harness_* tools, same key, no install. For n8n's MCP Client node and any HTTP-MCP agent, point the client at the URL and the tools appear immediately. The stdio ejentum-mcp npm package keeps working unchanged for Claude Desktop, Cursor, Windsurf, Claude Code, Cline, Continue, and other subprocess-spawning clients.
  • ejentum-mcp v0.1.11 published with a remotes[] block in server.json declaring the hosted endpoint alongside the npm stdio package. The Official MCP Registry surfaces both transports to its consumers (PulseMCP, downstream registries cascade automatically). npm description, plugin manifests, and integration files all updated to surface the dual-path framing.
  • Tighter recommendations: stdio install is via npx -y ejentum-mcp direct from npm. The earlier "one-click install via Smithery" pattern is no longer the recommended path; Smithery is still listed as a passive marketplace presence but our docs now lead with the hosted endpoint or the direct npm install.
  • See the updated MCP integration guide for the comparison table and per-client setup.

v1.1. May 2026

Added: official MCP server.

  • New product: ejentum-mcp, an MCP server exposing the four cognitive harnesses as MCP tools any agentic client can call (Claude Desktop, Cursor, Windsurf, Claude Code, n8n MCP Client node, custom Python or TS agents).
  • Tool names: harness_reasoning, harness_code, harness_anti_deception, harness_memory. Each takes a query argument and returns the same scaffold the Logic API returns directly.
  • Listings: Smithery (one-click install), Glama, mcp.so. MIT licensed; source on GitHub.
  • Backwards compatible: the MCP server is a thin wrapper over the existing v1 Logic API. No request/response schema changes. Existing API consumers, n8n workflows, and skill-files installs continue working unchanged.
  • Walkthrough: see the MCP integration guide for per-client install steps and the comparison vs HTTP Request and skill files.

v1.0. March 2026

Production launch.

  • 311 cognitive abilities across 6 reasoning dimensions (Causal, Temporal, Spatial, Simulation, Abstraction, Metacognition)
  • Two modes: single (1 ability per call) and multi (4 synergized abilities per call)
  • Response keys: single_ability and multi_ability
  • Canonical response sections: [REASONING INJECTION], [NEGATIVE GATE], [PROCEDURE], [REASONING TOPOLOGY], [TARGET PATTERN], [COGNITIVE PAYLOAD], [FALSIFICATION TEST]
  • Rate limiting: 100 requests/minute per key
  • Tier-based monthly quotas: Free (100 total), Ki (5,000/month), Haki (10,000/month)
  • Mode gating: multi mode requires Haki plan
  • Endpoint: POST /logicv1/

Stability guarantee

The v1 response schema will not change without a new version path. Field names, section delimiters, and JSON structure are frozen for v1.