LangChain

Integrate Intutic governance with LangChain v1.x agents.

LangChain is governed on two independent surfaces:

1. LLM egress — point the model client's base_url at the Intutic proxy (or launch under intutic exec). Every LLM call crosses the proxy and is governed like any other harness.
2. Local tool execution — LangChain tools run as plain Python callables/objects inside your process. No config or hook file can gate them, so the blocking gate ships SDK-side in intutic-clawde.

How it works

The langchain adapter is detected when pyproject.toml, requirements.txt, or uv.lock declares a langchain/langchain-core dependency. Like LangGraph, it writes a .env.intutic file with proxy base-URL env vars plus a comment block pointing at the SDK gate.

Detection also covers LangChain.js (the langchain npm package) for reporting purposes — but this adapter's env-writer is Python-only, matching the intutic_clawde.gate.adapters.langchain gate it points at. A JS/TS tool-call gate for LangChain.js is tracked separately.

Setup

1. Initialize Intutic

intutic init

✓ Detected harnesses:
  • langchain → .env.intutic

2. Route LLM traffic through the proxy

Source the generated env file, launch under intutic exec, or set base_url explicitly — same as LangGraph.

3. Gate local tool execution (SDK)

pip install intutic-clawde[langchain]

LangChain v1.x exposes a documented middleware veto point — AgentMiddleware.wrap_tool_call(request, handler) — which IntuticMiddleware implements:

from langchain.agents import create_agent
from intutic_clawde.gate import Gate, GateConfig, install
from intutic_clawde.gate.adapters.langchain import IntuticMiddleware

install(Gate(GateConfig()))

agent = create_agent(model, tools, middleware=[IntuticMiddleware()])

On deny, IntuticMiddleware returns a ToolMessage(status="error") carrying the [Intutic Governance] BLOCKED: ... message without calling the handler — the tool body never runs.

Pre-1.0 LangChain, or plain tool objects: use the framework-agnostic guard_tools(tools) helper instead — it already duck-types .func/._run on LangChain tool objects without importing langchain, and needs no LangChain-version-specific code:

from intutic_clawde.gate import guard_tools

tools = guard_tools([shell_tool, write_file_tool, deploy_tool])

Both paths funnel into the same Gate.guard(tool_name, tool_input) decision — the same four-tier evaluation (policy snapshot, SOP rules, image integrity, control-plane hook-gate) documented on the LangGraph page.

4. Trace attribution

from intutic_clawde.gate import intutic_headers

llm = ChatOpenAI(
    base_url="http://localhost:4000/v1",
    default_headers=intutic_headers(session_id=run_id, harness="langchain"),
)

What gets written

Same shape as LangGraph's .env.intutic — proxy URLs plus a pointer at intutic_clawde.gate.adapters.langchain.IntuticMiddleware.

What the adapter does NOT do

Same structural gaps as every SDK-gated framework — see LangGraph's "What the adapter does NOT do". In short: there is no sync-daemon hook file, argument-level gating requires wiring the SDK into your own agent code, and x-intutic-harness attribution is client-supplied, not authorization.

Config details

Property	Value
Harness type	langchain
Config file	.env.intutic
Detection	langchain/langchain-core in pyproject.toml, requirements.txt, or uv.lock (Python); langchain in package.json (JS/TS, detection-only)
Format	Shell environment variables
Write strategy	Atomic (write to .intutic-tmp, then rename)
Tool gate	SDK-side (intutic_clawde.gate.adapters.langchain.IntuticMiddleware, LangChain v1.x wrap_tool_call) — no sync-daemon hook file