Ship AI-written code
you can actually trust.
Dualzo is the AI development governance platform for GitHub. It turns a GitHub issue or a prompt into a human-approved, sandbox-validated pull request, checked against your repo's own tests, with every step audited. Pick the model that does the work: Claude, Codex, or Gemini.
No card required · no API keys to manage · 🎁 300 free signup credits (valid until Oct 31, 2026)
Connect a repo of any language. Dualzo indexes it and opens the PR for you.
Runs your repo's own install / lint / test / build in an ephemeral container.
Nothing runs until a person approves the plan. Signed links or the web view.
Every webhook, plan, approval, and PR is recorded, append-only.
Four steps. A human stays in the loop on the one that matters.
Write a prompt in-app, or drop the dualzo label on a GitHub issue to start a task from there.
It slices Project Memory for context and asks your chosen model — Claude, Codex, or Gemini — for an ordered, task-by-task plan.
Three tasks across the middleware and its tests.
Approve, reject, or tell it what to change. Approval is required — code never runs on a plan no one signed off.
Each task is validated in a sandbox. A PR is impossible to open while validation is failing — enforced in code.
Dualzo is the governance layer — plans, approvals, validation, audit. The model underneath is yours to choose, per repository or per task. Dualzo runs it for you — no API keys to set up.
Every plan is sent for approval by email and to your chat channels, with one-click Approve and Reject. The first person to act decides. All included free.
Every workspace member gets the plan, risk, and signed single-use Approve / Reject links.
Approval cards land in the channel you choose, with Approve and Reject buttons.
An Adaptive Card per plan, posted through a Teams Workflows webhook.
A card in your Chat space with the plan summary and one-click actions.
Same models, same repo. The difference is everything around the code. See how Dualzo compares with GitHub Copilot →
| Agent on its own | With Dualzo | |
|---|---|---|
| Plan before code | Starts editing straight away | ✓ Written plan with tasks, risk, and files expected |
| Human approval | Optional, ad hoc | ✓ Required before anything runs |
| Validation | Whatever you run locally | ✓ Your repo's own lint, typecheck, test, and build in a sandbox |
| Failing checks | Can still reach a PR | ✓ A PR cannot be opened while validation fails |
| Secrets | Runs with your local environment | ✓ Ephemeral container, no Dualzo secrets, network only for installs |
| Audit trail | Terminal history, if kept | ✓ Append-only log of every step |
One plan. No seat tiers, no usage surprises on Dualzo itself.
5 seats included · billed via Paddle · cancel anytime
AI model usage included · credit-based usage limit
AI code governance is the set of controls around AI-written code: a written plan before any change, human approval before anything runs, validation against the repository's own checks, and an audit trail of every step. Dualzo provides that layer for GitHub repositories.
With Dualzo it is the default. Every plan is sent to your team by email, Slack, Microsoft Teams, or Google Chat, and nothing runs until a person approves it with a signed, single-use link or in the web view.
Copilot helps a developer write code in the editor. Dualzo governs AI agents that work on their own: it turns a GitHub issue into a plan, waits for approval, validates the result in a sandbox, and opens a pull request that a human reviews and merges.
Any language. Dualzo detects the ecosystem, parses the manifest, and validates using the repository's own install, lint, typecheck, test, and build commands, so support does not depend on hand-written per-language analyzers.
No. Dualzo opens a pull request and stops. A human always reviews and merges. Nothing executes at all until a person approves the plan.
Claude, Codex, or Gemini — chosen per repository or per task. Dualzo runs the model for you, so there are no API keys to set up.
Each task runs your repository's own commands inside an ephemeral container with CPU, memory, and time limits, network access only during dependency install, and no Dualzo secrets in the environment. A pull request cannot be opened while validation is failing.
No. Your repository is cloned into an isolated working directory for the job at hand, and that directory is deleted when the job ends, including when it fails. Code is read only to plan, execute, and validate work you have approved.
Every step, from the issue to the approval, each task, each validation run, and the pull request, is written to an append-only audit log that cannot be edited. Entries are kept for 90 days and then pruned automatically.
One plan: Team at $119 per month with 5 seats included. Free credits are occasionally offered as a signup promotion — check the app for current offers. AI model usage is included in the plan, measured in credits, with a usage limit that refills automatically.
Team — $119/month, 5 seats included. Connect a repo, choose a model, and let Dualzo open the pull request while you keep the final say.