Metacontext

Every AI agent,
working from the same playbook.

AI agents forget your rules, follow outdated ones, and don't talk to each other. This setup gives them one approved, always-current source of truth, and keeps you in charge of it.

Get the free template → See it running in HighLayer ↗
See how it works ↓
Airtable base source of truth You PM agent Builder agent Reviewer agent
1

One place for every rule

Your rules, decisions and plans live in one Airtable base. Agents look them up there every time.

  • Nothing gets copied. Repo files like AGENTS.md and each task only point to the record in Airtable.
  • No drift. There's never a second version going stale in a doc or a chat.
YOUR REPO AGENTS.mdpoints to latest rules Task in a PRpoints to TASK-14 Review skillpoints to dev rules AIRTABLE BASE GS-01 ADR-01.01 TASK-14 DR-02.02 FD-01.02
2

Organized under Gold Standards

Gold Standards are the big themes for how your project is run. Every other record is filed under one and numbered within its family, like ADR-01.02.

  • Click any box in the diagram to see what it holds.
GOLD STANDARDProtect the user ADRDecisions FDPlanned features REQRequirements DRDev rules
3

Always current

Each record only says what's true now. When something changes, the record gets updated and its version goes up.

  • No pile of old versions for an agent to get confused by.
  • Agents never read old or replaced thinking.
ADR-01.01 · Decisionv3
Sign-in uses email magic links.
Approved by you
Other setups
Sign-in uses passwords
Sign-in uses Google only
See new doc for sign-in
This setup

One record. One version that counts. Watch it update in place.

4

You approve. Agents follow.

You're the only approver. Agents only act on records you've approved, and new ideas wait as future decisions so they don't creep into current work.

  • Try it: approve a record and watch it become available to the agents.
Waiting for you
Agents can use
5

How the work runs

From the moment you mark a task Ready to the moment it ships, every agent works from the same approved rules.

Every night, a Contradiction Agent checks the base. If two rules conflict, it flags them for you before any agent gets confused.
Why it matters

You set direction once. Every agent stays aligned.

It's governance and memory that sit outside any one AI tool. Claude, Codex and any agent you add later all read the same approved, current truth.

No re-explaining

Agents pull what each task needs, so you stop pasting the same context into every chat.

Every action traces back

Each line of work points to a decision you made, under a Gold Standard you set.

You stay in control

You own the what and the why. The agents handle the how.