Preview · measurement is disabled for this view.
HOW IT WORKS
Ask your agent. It uses Weft.
Give your agent a change to make and the checks it must pass. It uses Weft to record the task, run the work, and keep the result with your project.
- You describe the result.What should change? How will you check it?
- Your agent runs the work.Local workers follow the workflow and run the checks.
- You keep the important choices.Your agent asks when a significant decision needs your answer.
Add a CSV export.
- Owner
- Implementing agent
- Done when
- Export passes the repository’s checks
- Scope recordedOne bounded change.
- Check → revisionThe failed check sends work back.
- Build + tests passThe result is ready for acceptance.
Illustrative local task · your agent does the work.
- Set the scope.
- A failed check sends work back.
- Revise, then verify the result.
- Start the next task with context.
CHECK THE RESULT
A failed check means there’s more to do.
Make your build, tests, and other required checks part of the workflow. If a check fails, the affected task waits while the agent fixes the problem.
How do checks control the workflow?
A gate is a check attached to a step in your workflow. When it is required, that step cannot advance without a passing result.
Weft OSS includes structural checks and build and test verification. Use your repository’s tests to check the behavior you care about. A gate can be as simple as running a CLI tool or shell script you already use. Your coding agent can write a small plugin wrapper when you need a new check.
AI review is available through an added review provider, including optional Weft Cloud review.
- Build the change.
- A check finds a problem.
- Revise the work and check again.
- Accept after the checks pass.
KEEP THE ANSWER
Explain a decision once. Give the next agent the reason.
Ask your agent to search project memory before a significant change. If there’s no answer, it asks you and records your decision and why you made it.
A later session can retrieve that answer—even when you switch agents or models.
See how a decision carries forward →- Decision
- Keep original timestamps.
- Reason
- The export must preserve the source history.
Illustrative handoff · agents record and retrieve decisions.
- No prior answer: ask the person.
- The person supplies the answer.
- Save the decision and its reason.
- Retrieve the answer for later work.
MAKE IT YOURS
Change the process without rebuilding the machinery.
Start with the included software development workflow. Ask your coding agent to edit the YAML: add a stage, change the order, or choose where a check must pass.
Reuse that process for the next task. For a different kind of work, ask your agent to write a workflow of your own. You can build a plugin when you need new tools or checks.
Open the workflow file →- Design
- Implement
- Verify
- Design
- Implement
- Your review
- Verify
Illustrative flow · customize stages and checks in your project.
- Start with the included software development flow.
- Add a review stage to your workflow definition.
- Run the next task with your revised process and the same engine.
CHOOSE WHO DOES THE WORK
A smaller job can use a smaller model.
Choose a lower-cost model for a well-defined change and a stronger model for harder work. Keep the acceptance checks whichever model does the work.
You or your agent make that choice, using your own model accounts.
See an example of model selection →Lower-cost implementing model
Stronger reasoning model
Illustrative assignments · you or your agent choose the model.
START SMALL
One change. Your agent. Your checks.
You don’t need to design a factory first. Start with the included workflow, then adjust it as you learn what your work needs.
Tasks, project memory, workers, and checks run locally. Free and open source under Apache 2.0. No Weft Cloud account or subscription required.
When would I add Cloud? →GETTING STARTED
- Set up Weft for your agent.Connect through MCP or use the CLI.
- Bring a small change.Tell your agent what success looks like.
- Run it, check it, reuse the process.Keep the work record for the next task.
Setup and launch availability
Once public installation is available, install the Weft binary and register it with your MCP client. Guided setup supports Claude Code, Cursor, and Codex. Other agents can use Weft through their MCP integration or the CLI.
# After public launch and installation:
weft install --client claude-code
# or:
weft install --client codex
weft init
Restart your assistant, ask it to create a Weft project and a small task, and start with checks you already trust. Your agent remains the interface; you can use the CLI directly whenever useful.
OSS launch is being prepared. Public installation and repository access will be announced here. You can explore the process now and ask for launch updates.
FOLLOW THE OSS LAUNCH
Give your agents a process at launch.
Get an email when Weft OSS is available. Start with one project, your existing coding agent, and checks you already trust.