Not another tool. The layer above them.
Your engineering organization already runs on developers, AI agents, contractors, and delivery partners. Foreman is the control plane that coordinates all of it — it coordinates the workflow, enforces the required gates, and preserves human release authority.
Stop deciding which AI tool to prompt next.
Foreman coordinates the work across connected coding agents, independent reviewers, your verification systems, and human release authority. Define the outcome once. Foreman coordinates the delivery path.
Work requirement
1 requirement
Independent review of candidate a52819de
Connected capabilities
3 connected
Coding agents available in this environment
Policy + eligibility
2 eligible
Cleared for the repo · did not author the work
Foreman Intelligence
Ranked
Execution history on similar work
Assignment
1 selected
OpenAI Codex · independent reviewer
Policy defines what is allowed. Intelligence helps choose among what is allowed.
One place where the work is governed.
A control plane doesn't do the execution — it decides what should happen, who is cleared to do it, and what must be true before anything ships. Foreman holds nine responsibilities so your tools don't have to agree with each other.
Work intake
Business intent shaped into scoped, reviewable work.
Policy
What is allowed, and who is cleared to do it.
Worker routing
Which eligible builder or reviewer gets which work, and why.
Independent review
A reviewer that did not author the work challenges it.
Independent verification
Your testing, security, and CI systems inspect the candidate.
Evidence
Results and findings, recorded against the exact candidate.
Authority
Where a human decision is required.
Release readiness
Whether the evidence satisfies your release policy.
Intelligence
Helps choose among eligible options — never overrides policy.
Builders build. Inspectors inspect. Foreman governs.
Foreman separates implementation from verification. Builders produce candidates. Independent inspectors evaluate them. Foreman applies policy across both.
Your organization can continue using the QA, security, and CI systems it already trusts.
- Builders · implement
- AI coding agents produce a candidate — the exact work to be evaluated. A builder never certifies its own work.For example Claude Code, OpenAI Codex, Muse, GitHub Copilot, Cursor.
- Independent AI review
- A separate agent that did not author the candidate reviews the code and decisions, and files findings.
- Independent verification
- Your testing, security, and CI systems inspect the same candidate and return results. They run in your environment, on your accounts.For example BrowserStack, Snyk, mabl, Tricentis, QA Wolf, Customer CI.
- Human decisions
- Agents can ask. Foreman can recommend. Your team decides — approve, reject, or request remediation.
Foreman collects the evidence, routes findings, and determines release readiness under your policy. Inspectors cannot approve a release.
Verification providers remain separately owned and billed by your organization. With Private Runner, those provider credentials remain in your environment. Foreman receives verification results and evidence—not the provider credentials themselves.
The human-facing interface to Foreman Intelligence.
Ask Foreman lets a delivery lead question the control plane directly. It reads the same work state, eligibility, and evidence Foreman uses to govern delivery — and answers with the reasoning behind it.
- What should happen next?
- Who is eligible to do this work?
- What is blocked?
- What testing is still required?
- Why is this candidate not ready?
- What requires my decision?
How Ask Foreman is bounded
Advisory only- Evidence-grounded
- Answers cite the work items, candidates, and evidence behind them.
- Workspace-scoped
- It only sees the workspace you are signed in to.
- Advisory
- It recommends. Policy and people decide.
- Read-only
- It cannot approve, deploy, merge, or release on its own.
Ask Foreman cannot approve, deploy, merge, or release on its own. Those decisions stay with policy and people.
Foreman Intelligence may use approved model and research capabilities behind the scenes. They inform the answer — they never take over the workflow.
Intent enters. Governed delivery comes out.
The architecture starts before execution. Business intent is shaped into work, routed through the control plane to eligible capabilities, proven with the evidence it requires, and stopped at a person for release.
- 01
Business intent
An outcome, requirement, ticket, or problem.
- 02
Work shaping
Requirements, constraints, and evidence needs.
FOREMAN CONTROL PLANE
- Work intake
- Policy
- Worker routing
- Independent review
- Independent verification
- Evidence
- Authority
- Release readiness
- Intelligence
- 04
Delivery capabilities
Eligible connected AI capabilities, with human decision points preserved.
- 05
Independent review + verification
Separate AI reviewers and your own testing, security, and CI systems produce the evidence the work requires.
- 06
Human release
Nothing ships without a human decision.
- 07
Delivery intelligence
Execution history feeds the next assignment.
Built around your stack. No rip and replace.
Foreman coordinates the tools you already trust. Keep your agents, your source control, your QA and security systems, and your deployment pipeline — Foreman governs how they work together.
AI execution
Approved coding agents run under policy and routing, on your own provider accounts.
Source control
Candidates, branches, and evidence tracked to the exact SHA.
Design
UI and product-design work routes to a connected design capability.
Verification
The QA, security, and CI systems you already trust inspect the candidate. Separately owned and billed by you; Foreman coordinates their results.
Deployment
Release happens on your existing pipeline.
Private execution
Work runs inside your environment, next to your code.
Every plan connects Foreman to your own provider accounts. Foreman treats agents as interchangeable capabilities — you can add, remove, or reassign them without changing how work flows.
See the control plane on your own work.
We'll walk through how Foreman would coordinate your agents, your teams, and your release process — using the way you already build.