hexFoundry · Collaborator

AI that finally remembers what your team decided.

Collaborator turns code review into the training loop. The comments you already write become durable lessons your agents inherit — so you stop re-explaining the same conventions every session.

01 The problem

AI coding today is amnesiac. Every session starts from zero.

// again

You correct the same mistake you corrected last week. The model has no memory that you ever did.

// re-explain

Your conventions — error handling, naming, which library you standardized on — get re-typed into the prompt for the hundredth time.

// re-litigate

Architectural decisions the team argued through months ago get quietly reopened, because nothing carried them forward.

The review comment you wrote last Tuesday is gone. Your team keeps paying interest on knowledge it already earned.

02 How it works

Review becomes the loop that teaches the codebase.

Humans and AI agents are symmetric workers on the same repo. You review the diff exactly like you always have — and that review compounds.

STEP 01

Comment on the diff

Engineers review pull requests the way they always have — inline comments on the exact lines. No new ritual, no separate knowledge base to maintain.

STEP 02

Lessons are distilled and voted on

When work closes, related comments are distilled into durable, org-scoped lessons — each one traceable back to the comments that produced it, and ranked by your team's own up and down votes.

STEP 03

Every future agent inherits them

Each new agent task opens with your team's top lessons injected into the prompt. The codebase teaches the AI what this team actually cares about — before it writes a line.

STEP 04

The loop tightens

Corrections get made once, not weekly. Good lessons rise, stale ones get voted down, and the quality floor moves up on its own. It compounds.

03 What's under it

Everything is tracked, so review gets cheap.

Every action writes to an immutable ledger. That's what lets Collaborator trust its own history — and lets you stop watching the boring parts.

lessons.compound

Lessons that compound

Distilled from real review comments, ranked by votes, scoped to your org. Every closed task can leave the next one a little smarter.

workers.symmetric

Humans and agents, one repo

People and AI agents are the same kind of worker: they open branches, push diffs, and get reviewed identically. No separate lane for "the AI code."

comments.dispatch

Comments that dispatch work

An inline review comment can hand the fix straight to an agent. The thread stays where the context is — on the exact lines you were reading.

approvals.hashed

Content-hashed approvals

Every diff block is content-hashed. If an approved block changes later, it automatically resurfaces for re-approval — nothing slips through silently.

gates.merge

Check gates at the merge boundary

Coverage and tests run at merge time, not as advisory noise. A failing gate can hard-block a bad merge outright — the rule is the wall.

containers.sealed

Sealed throwaway containers

Untrusted code — including anything an agent generates — runs in sealed, single-use containers. Nothing it does reaches back into your infrastructure.

coverage.13

Coverage across 13 ecosystems

Test and coverage instrumentation spans thirteen language ecosystems, so the gates mean the same thing in your polyglot monorepo as in a single service.

ledger.immutable

Immutable event ledger

Every action — comment, approval, merge, agent run — is an append-only event. You can replay exactly who or what changed a line, and why.

tenancy.rls

Per-org row-level security

Isolation is enforced in the database with per-org Postgres row-level security — not just in application code you have to trust to get it right every time.

04 The output

It lands on one thing: developer output.

More shipped, less re-explaining. Not because anyone typed faster — because the review you were already doing stopped evaporating.

once
A convention gets corrected a single time, then it's carried into every future task automatically instead of re-typed weekly.
1→1
Every lesson is traceable back to the exact review comments that produced it. Nothing is a black box you have to take on faith.
0
Silent re-approvals. Content hashing and merge gates mean a change can't sneak past the review it already passed.

Engineers spend their attention on the decisions that matter. The boring code writes itself.

Get started

Stop re-teaching the same lessons.

Bring Collaborator to your repo and let review start compounding. Early access is rolling out to engineering teams now.

Re-read how it works