Guides
Pick the integration path that matches your stack, then dive into the substrate primitives.
Getting started
Read these first. The pain-point map and a five-minute install-to-run walkthrough.
Integrations
Mount Entorin under whatever loop you already use — SDK or hand-written.
- Claude Agent SDK
Mount Entorin under the Claude Agent SDK via httpx transport, wrapped MCP sessions, and a per-run sandbox.
- Codex CLI / Codex SDK
Same shape as the Claude SDK path, but for the OpenAI wire format — Chat Completions and the newer Responses API both supported.
- LangGraph
Wrap each node and drive the graph through HITL — substrate primitives flow through the same RunContext / EventBus / Ledger as every other path.
- Bare loop, no SDK
If you don't want a framework, you don't need one. A 50-line while loop inherits the full harness.
Primitives
Substrate primitives every harness reinvents — memory, lease, policy, fanout, retrieval, skills.
- M6 primitives — memory, lease, policy, fanout
Four protocol-first primitives every harness reinvents. Backends are pluggable; the shipped reference adapters get out of the way as soon as you have a real backend.
- M7 primitives — retrieval, skills
Two more protocol-first substrate primitives. Same backend-pluggable discipline as M6.
Trace replay
Saved traces become regression fixtures. Assert invariants without re-running the model.