AI Code Worker

AI that changesyour code —under your rules.

A standalone, multi-tenant service that runs AI coding tasks against repositories you have explicitly allowlisted, inside isolated per-tenant workspaces. Your product calls one HTTPS API, watches a live transcript, stops or steers the run, and receives a signed terminal callback. The caller proposes; policy disposes; humans merge the pull request.

Allowlisted repos · Isolated workspaces · Signed callbacks · Humans merge PRs

What the service guarantees
3Task kindsfix · review · investigate
SSELive transcriptstop & steer mid-run
HMACSigned callbacksverified before parse
v1Frozen API contractOpenAPI 3.1 + test vectors
The problem

AI coding agents are trivial to demo and genuinely dangerous to operate.

  • 01Agents run with broad credentials against every repository they can reach — one careless prompt touches production code.
  • 02You are handed a diff with no transcript, so nobody can reconstruct which commands produced it or why.
  • 03A long-running agent that dies quietly leaves the calling system waiting forever, with no terminal signal to act on.
  • 04Every product that wants AI code changes rebuilds the same sandboxing, queueing, and webhook plumbing from scratch.
Agent bolted on vs. a governed service

What changes when AI code execution becomes its own controlled system.

An agent bolted onto your app

  • Broad tokens and ambient repository access, with hope as the only control.
  • Runs on the application server, competing with user traffic for CPU and memory.
  • Fire-and-forget: no way to halt a run that has gone wrong or nudge it back on course.
  • A webhook any caller can forge, because nothing verified the signature before parsing the body.
  • Agent logic welded to one product's data model — unusable for the next product.

AI Code Worker

  • A per-tenant repository allowlist; write capability is credential-scoped and read-only kinds never receive a write token.
  • Isolated per-tenant Kubernetes workspaces — one run cannot reach another tenant's code or secrets.
  • A Server-Sent Events transcript with stop and steer available while the task is still running.
  • HMAC-signed terminal callbacks verified before parse, failing closed, backed by a durable retry outbox and manual redelivery.
  • One product-agnostic HTTPS API — the second consumer integrates without a line of change in the service.
Platform capabilities

Every control you would have to build yourself, already in the contract.

The service is deliberately product-agnostic: consumers integrate over HTTPS with a scoped API key, and no consumer-specific logic lives inside it. Control plane on Python 3.12, FastAPI, and Postgres — the task queue and callback outbox are Postgres-backed, so there is no Redis to operate. Workspaces run on Kubernetes.

Three task kinds

code_fix, code_review, and investigate. Only code_fix can write, and it must be given an explicit write target and branch; read kinds are issued no write token at all.

Policy-allowlisted repositories

Each tenant carries a repository allowlist managed over the API. The caller proposes a repo and branch and policy disposes — a request outside the allowlist is refused, not attempted.

Isolated workspaces

Every run executes in its own per-tenant workspace that holds no platform credentials, so a compromised run yields nothing reusable against the service or another tenant.

Live transcript, stop & steer

Stream events over SSE using short-TTL stream tokens, backfill the transcript at any time, stop a run outright, or queue steering guidance applied at the next turn boundary.

Signed terminal callbacks

Every task ends in an HMAC-signed callback — task completion plus pull-request lifecycle events — or in a synthesized failure. Silence is never a valid outcome.

Durable delivery outbox

A Postgres-backed outbox tracks each delivery through pending, delivered, failed, and exhausted, with retries and manual redelivery when your endpoint was down.

Humans merge pull requests

The worker opens pull requests and never merges them. Review stays exactly where your team already reviews, and merge authority stays with people.

Reconstructable audit trail

Task listing, event backfill, delivery history, and tenant self-inspection mean any run can be reconstructed after the fact — including the ones that failed.

Product-agnostic by test

The contract test harness always exercises a second, synthetic consumer shape alongside the real one, so the API cannot quietly specialise to a single caller.

Where it fits

Any product that wants code changes it can defend.

The same API serves very different callers, because the service knows nothing about their domains — only about tasks, repositories, policy, and callbacks.

Feedback & issue reports

From bug report to pull request.

A product's feedback module hands a triaged issue to the worker and receives a pull-request link on a signed callback — with the full transcript attached to the report.

Code review at scale

A reviewer on every branch.

code_review runs against allowlisted repositories and returns findings with no write capability whatsoever — useful precisely because it cannot change anything.

Incident investigation

Read-only root-cause runs.

investigate explores a repository to answer a question and returns findings. No branch, no write target, no token — the safest kind to hand to on-call.

Platform teams

One worker, many products.

Internal products each get their own tenant, allowlist, and keys, and integrate over the public API only — no shared database, no bespoke branch per consumer.

How we deliver

From contract to a worker your product can call.

01

Contract & scoping

We agree the task kinds, repositories, and callback endpoint. The frozen v1 OpenAPI contract and the published signature test vectors are the specification you build against — not a moving target.

02

Tenant & credentials

Scoped API keys are issued per client alongside a paired callback signing secret. Keys are shown once, stored only as hashes, and split by scope — application keys stay separate from policy administration.

03

Allowlist & policy

Repository allowlist, permitted write targets, branch rules, and hard wall-clock caps are configured per tenant before the first task is ever created.

04

Integration & contract tests

You verify your side against the shipped harness — consumer fixtures, deliberate negatives, and signature vectors — so the integration is proven rather than assumed.

05

Operate & evolve

Task, event, and delivery history give you the operational view. Credentials rotate by overlap, failed callbacks redeliver on demand, and new task kinds land additively — consumers tolerate unknown kinds by contract.

Inside the service

A task you can watch, stop, and audit — not a black box.

Representative views of the two surfaces that matter: what your product shows a user while a task runs, and what your platform team watches across tenants.

AI Code Worker
Task · code_fixRunning · streaming
The invoice importer crashes when a tax line is missing. Fix it.
Task accepted and running in an isolated workspace. Streaming the transcript — you can stop or steer at any point.
task_7f2Kq · code_fixAPI · v1
Repositoryerp-connector
PolicyAllowlisted
Write targetfix/tax-line-null
StatusRunning

Your users file an issue; the transcript streams back in your own UI.

codeworker.promaxgulf.com / tasks
Tenants › Tasks › Control centre
Task control centre
Multi-tenant · allowlisted repositories
Transcript · Live
Running now
4
isolated workspaces
PRs awaiting review
7
humans merge
Callbacks delivered
100%
outbox drained
Policy refusals
2
repo not allowlisted
Task lifecycle · todayWorkflow automation · live
Queued148
Running141
Terminal139
Callback delivered139
Recent task eventsAudit trail
  • 11:06task_7f2Kq · code_fix opened PR #418 on erp-connectorOK
  • 10:52task_4m8Lb · investigate returned findings — no write token issuedOK
  • 10:44task_2c9Xr · code_fix refused — repository outside tenant allowlistWARN
  • 10:31Callback for task_9d1Vt redelivered after consumer endpoint recoveredOK

Runs, policy decisions, and callback deliveries across every tenant.

k8s
Isolated per-tenant workspaces
OpenAPI
3.1 contract, frozen v1
Postgres
Queue & outbox, no Redis
Common questions

What engineering leaders ask first.

No. Each tenant has an explicit repository allowlist. The caller proposes a repository and branch, and policy disposes — anything outside the allowlist is refused rather than attempted. Only code_fix can write, and it must be given a write target and branch; code_review and investigate are issued no write token at all.

Next step

Give your product AI code changes — without handing over your repositories.

Tell us the product, the repositories, and which task kinds you need. We will propose the tenant model, policy allowlist, callback design, and a first integration your engineers can verify against the shipped contract tests before anything touches a real branch.

Saudi-based delivery · Contract-first · Humans merge every PR

ProMax TechPro Max Gulf

Premier technology partner delivering enterprise-grade ERP, AI, software, and hardware solutions across Saudi Arabia.

Quick Links

Services

© 2026 Pro Max Gulf. All Rights Reserved.info@promaxgulf.com