Watchlight Developer Edition

Govern an AI agent in five minutes. One install, zero infrastructure, same API as production.

pip install watchlight        # Python 3.9+, prebuilt wheels
npm install @watchlight/sdk   # Node 18+, the compiled engine comes with it

Watchlight puts a policy decision point in front of every action your agents take. It authorizes each tool call and records a value-free audit trail.

The Developer Edition runs the real authorization engine in-process — no server, no database, no signup. When it says Deny, it denies for the same reasons production does.

from watchlight import govern, Denied

# Permit ONLY "research". Fail-closed: everything else is denied.
govern.allow('permit(principal, action == Action::"research", resource);')

@govern.tool(intent="research")
def web_search(query: str) -> str:
    return f"results for: {query}"

@govern.tool(intent="transfer")           # governed, but no policy permits it
def transfer_funds(to: str, amount: int) -> str:
    return f"sent ${amount} to {to}"      # never runs — denied first

web_search("watchlight docs")             # ALLOW → the body runs
transfer_funds("mallory", 1000)           # DENY  → raises Denied, before the body

That DENY is the product. The transfer_funds body never ran, and the decision is already on disk in .watchlight/audit.jsonl.

→ Quickstart — the full program, its output, and the live console.

What you can govern

	Python — pip install …	Node — npm install …
A tool you wrote	watchlight → @govern.tool(intent=…)	@watchlight/sdk → govern.tool(fn, { intent })
A framework agent	watchlight[langgraph] · [pydantic-ai] · [claude-agent] · [deepagents]	governedHooks() for the Claude Agent SDK, governTool() for LangChain / LangGraph.js
An MCP server	watchlight-mcp — a policy enforcement point in front of any MCP server	point any MCP client at that same PEP
Your own app	govern.authorize(…)	await govern.authorize({…})

pip install "watchlight[all]" is one install for the whole Python lane, and every Python example in these docs runs after it. On the Node lane @watchlight/sdk is the only install you need — see TypeScript / Node.

Both lanes write .watchlight/audit.jsonl, and watchlight dev tails it:

watchlight dev            # → http://localhost:7000

Who is acting, and on whose behalf

A governed call answers two separate questions, and that is what separates a policy from an if.

Question	Where it goes
On whose behalf does this run?	principal — the subject, e.g. User::"db:4412"
Which runtime is acting?	context.actor, set by the SDK from the agent name
Through whose delegation?	context.actor_chain
Under what narrowed authority?	govern.scope(tools=[...])

The SDK sets context.actor itself and refuses a caller-supplied value that disagrees, so a policy can trust it. → The identity model

A note on identity

The Developer Edition authorizes the principal you assert — the agent you construct the governor with, or the Watchlight-Agent-Id on a governed MCP request. It does not cryptographically prove the caller. On your own machine, running both sides, that is the right trade.

Bind any non-loopback listener behind something that authenticates the caller — a reverse proxy doing mTLS or OIDC, or the Enterprise plane. Identity hardens as you grow without changing your policies.

The license

Everything you write against is open and Apache-2.0: the govern decorator, scope attenuation, the CLI and watchlight dev, the framework plugins, every example.

The decision engine ships compiled (watchlight-engine, and the watchlight-mcp runtime) under the Developer Edition license — free, including in production and commercially. A commercial license is needed only to re-offer the engine itself as a hosted authorization service. To govern agents across a fleet from one control plane, talk to us about Enterprise.

Where to go next

• Quickstart — your first ALLOW / DENY.
• Using the governor — where the governor goes in a request handler, a worker, a test.
• How policy works — the Cedar model, in five minutes.
• Govern an agent you already have — LangGraph, Pydantic AI, the Claude Agent SDK.
• Developer Edition vs Enterprise — what the control plane adds, and when to reach for it.