Decision intelligence for AI agents

Ask what your agents decided.

The decision of record for AI agents.

Every decision your coding agents make — recorded, governed, and searchable. Ask in plain language, scoped by team, person, repo, or day.

Install into any repo
# Install into any repo
npx precedent-cli init
Not a memory layer

Memory recalls. Precedent binds.

A memory layer

Stores facts and past context so an agent can recall them. Useful — but recall is optional, and nothing stops the next agent from quietly deciding otherwise.

The decision layer

Records a decision with its authority and scope — then enforces it. A guardrail isn’t just remembered; it blocks the PR that breaks it and cites why.

The loop

Three moves, inside the agent’s workflow.

Consume

In the agent's loop

Before it acts, the agent asks whether a decision already covers the change — and records new ones.

Enforce

At the merge gate

The same decisions block a violating PR and auto-merge a compliant one. Enforcement runs in your repos.

Govern

From this console

Browse and query decisions across org, team, person and repo. Promote what should bind.

In practice

An agent tries to break a rule — and gets stopped.

agent · merge gate
# Agent: add Redux for dashboard server-state caching
$ precedent govern --title "add Redux for server-state"

⛔ BLOCK — denied: violates a guardrail (E2)
   cites the decision below · not a reviewer, days later
The decision of record
Redux is banned for server-state caching; use React Query.
E2guardrailorgweb/**
Retrieved the guardrail before the commit; the merge gate blocks the PR and cites the decision — not a reviewer, days later.
Works with your agents

One command. Every harness.

“npx precedent-cli init” wires Precedent into every agent you run — retrieve, govern, record — without overwriting your existing rules.

Claude CodeCursorCodexGitHub CopilotWindsurfGemini CLIAiderZed
Start

Give your agents a decision of record.

Open source. Runs on the free tier. Enforcement runs in your repos.