CodeLynq turns a description of work into a pull request in your own repository. You stay in control at every step: you decide what gets built, you approve the plan before any code is written, and you review and merge the result in your git host as usual.
The flow at a glance
- A request comes in. You describe what you need in the new-task chat, in a task form, from meeting notes, from an alert or through an external AI client.
- It becomes a task. A task has a title, a description, a type (feature, bug fix, refactor, chore or troubleshoot), a priority and - if needed - the repositories and branches it applies to. See Tasks.
- The task gets a spec. CodeLynq proposes a technical solution design (TSD): how it interprets the request, the technical approach, affected components, risks and test strategy. It comes with the decisions it wants you to confirm, like a technical lead would ask; you answer them, the spec is updated, and you approve it.
- You approve the spec. Nothing is built until a TSD is approved. See The spec (TSD) and approval.
- A run does the work. A development run checks out your repository, makes the change, runs your checks and reviews its own work. See Runs and Quality gates.
- You get a pull request. The run opens a pull request in your git host - one per repository or branch it changed. You review it there.
- Feedback and fixes. If CI fails or a reviewer asks for changes, CodeLynq can work the feedback into the same pull request. See The CI feedback loop.
- Merge and deploy. You merge as usual. If you set up deploy targets, CodeLynq can trigger your deployment as well. See Deployments.
What CodeLynq knows about your code
Each project holds context that every spec and run reads: your architecture, conventions, test strategy and constraints, plus facts CodeLynq learns about the codebase over time. The better this context, the better the results. See Projects and Codebase memory.
Larger pieces of work
- An epic splits a bigger goal into tasks that run in dependency order.
- For a brand-new application, a scoping interview turns your idea into an architecture and a build plan.
- When something is broken and the cause is unclear, a troubleshoot task investigates first and turns the diagnosis into a fix task.
Staying in control of cost
Runs use AI models, and bigger tasks cost more. You see an estimate before you start a run, and your organization can set limits so expensive runs need approval first. See Approving expensive runs and Billing and credits.
Ready to try it? Start with Getting started.