Open Source · MIT License

The cognition layer that grows with your team.

Your AI understands before it codes — and your team's knowledge compounds every commit. Precept makes the agents you already use grasp what your team has decided, what a change will break, and what to reuse, then remembers it for next time.

Get started
Paste this to your agent
Set up Precept on this machine so my coding agent consults it before writing code.

1. Install the CLI: uv tool install --force git+https://github.com/qorstack/preceptai.git
2. Make a dedicated folder (e.g. ~/precept), cd into it, and run precept quickstart — it scaffolds the stack there, starts Postgres + the dashboard, registers the MCP server globally with Claude Code, and installs the /precept slash commands.
3. Verify the dashboard answers at http://localhost:9080 and claude mcp list shows precept.

If any step fails, read the error, fix the root cause, retry, and report what you changed.
Dashboard → localhost:9080 · .env + docker-compose.yml land in that folder; the MCP server & slash commands install globally (every repo).
The problem

AI agents generate confidently. They rarely understand.

They reinvent helpers that already exist, miss the payment flow three files over, and ship changes your team explicitly decided against last quarter. Markdown docs don't fix this — agents skip them. Precept makes understanding a required tool call, not a suggestion.

Intent → Impact → Risk

Every request runs through a rule-based pipeline that names the domain, the blast radius, and a binding decision — no LLM, fully reproducible.

Reuse before you create

Surfaces the exact existing assets and conventions to follow, so agents import instead of reinventing.

Binding risk gates

ASK pauses for a human. REJECT stops cold. Code is validated before it's written.

Where it fits

Not another agent. The layer your agents are missing.

Coding agents are great at producing code. None of them make sure it's the right code for your system, or that it respects what your team already decided. Precept sits in front of them over MCP and makes understanding mandatory.

Today's AI coding agents
Generate & edit code, fast
Personal, per-session memory
One developer, one machine
Optimize for speed of output
Precept · the cognition layer
Enforces understanding before any code is written
Team-shared, human-approved memory — git-synced
Every repo, any MCP client — one shared brain
Optimizes for correctness & team alignment

Point it at the agents you already use — Claude Code · Cursor · Cline · Hermes Agent · any MCP client. It doesn't replace them. It makes them think first.

How it works

Your agent asks. Precept answers with cognition.

Before touching code, the agent calls analyze_intent over MCP and gets a structured report it must respect. Here's what it sees for a real request.

In Claude Code
/precept add a refund endpoint to /payments
What the agent must do
1. analyze_intent()      # first, always
2. get_reusable_assets()  # reuse, not reinvent
3. assess_risk_in_context()
4. validate_generated_code() # before write
Domain
payment · action: create
Requires
Idempotency keyAudit logAmount validation
Impact
3 services, 7 files — touches the settlement worker & webhook retry path.
Reuse
payments/refund_service.py already exists — extend it, don't recreate.
Decision
Ask — human confirmation
Team knowledge

One brain your whole team — and every agent — shares.

When someone states a team decision, Precept captures it, a human approves it, and it git-syncs to everyone. The next time any agent in any repo touches that domain, the decision is right there in the report. Personal agent memory can't do this — it isn't governed, approved, or shared.

Save a team decision
precept memory decide payment \
  "Use idempotency keys" \
  --body "All payment calls require an
  idempotency key. No exceptions."
Workspace · Payment ✓ Approved · git-synced

Use idempotency keys

All payment calls require an idempotency key. No exceptions. Surfaced automatically on every payment-domain request.

Decided by @team · recalled in 14 sessions
Human-approved memory. AI proposes, a person ratifies. Nothing trusted lands unreviewed.
Workspace-scoped & multi-repo. One knowledge home, auto-linked across every repo in the product.
Audit log. Every recall, save, and decision is recorded — who, what, when.
Approval gates. High-risk changes wait in a queue until a human approves or rejects.
Tool-agnostic. Any MCP client — Claude Code, Cursor, Cline — calls the same brain.
Ship code your team actually agreed on

Make your agents understand before they act.

Open source. MCP-native. Two commands to a governed, shared cognition layer.