# Hakama

Hakama adds policy contracts, review gates, and evidence trails to Claude, Codex, and Gemini workflows so engineering teams can ship agent changes confidently.

Canonical: https://hakama.ai/

## Services


### Intent Contracts

Turn a natural-language request into explicit scope, constraints, acceptance criteria, and allowed execution boundaries.

- Task scope
- Risk profile
- Acceptance checks


### Review Gates

Check agent output against policy, repository rules, tests, dependencies, and documentation expectations before handoff.

- Diff inspection
- Policy checks
- Test evidence


### Audit Trails

Keep prompts, contracts, diffs, decisions, and validation output together so teams can review what changed and why.

- Run history
- Decision records
- Team review


## Selected Work


### Contracted Prompts

Category: Intent

Translate loose agent requests into scoped work orders with constraints, risk notes, and acceptance criteria.


### Guarded Changes

Category: Review

Check code changes, package edits, generated docs, and test evidence before a developer accepts the work.


### Decision Records

Category: Evidence

Keep the original request, agent output, validation results, and reviewer decision in one durable trail.


## Latest Articles

- [How AI coding agents quietly break your codebase](https://hakama.ai/how-ai-coding-agents-break-your-codebase/index.md): AI coding tools fail in recognizable ways: polished wrong output, claims ahead of evidence, scope drift, stale context, and decisions disguised as options. Catch the pattern before it reaches a commit. Canonical: https://hakama.ai/how-ai-coding-agents-break-your-codebase/

## Comparisons

- [Hakama vs GitHub Copilot](https://hakama.ai/compare/hakama-vs-github-copilot/index.md): GitHub Copilot helps your team write code. Hakama decides whether the change it produces is allowed to ship: scope, approvals, evidence, and a signed record. Most teams run both. Canonical: https://hakama.ai/compare/hakama-vs-github-copilot/

## Contact

[Request a pilot](https://hakama.ai/request-a-pilot/index.md) with Hakama. Canonical: https://hakama.ai/request-a-pilot/


## FAQ


### What does Hakama do?

Hakama adds a control layer around AI coding agents by turning requests into explicit contracts, checking output, and preserving evidence for review.


### Who is Hakama best for?

Hakama is best for engineering teams experimenting with Claude, Codex, Gemini, and similar coding agents that need clearer policy and review boundaries.


### How does a pilot usually start?

A pilot usually starts by mapping agent workflows, defining review gates, and choosing the evidence teams need before accepting generated code.


### Does Hakama replace code review?

No. Hakama makes code review easier by keeping intent, constraints, validation output, and agent changes together before a human makes the final call.