What is Kodel
Kodel is a platform for running software work through a process your team writes down. A task does not just sit in a tracker waiting for someone to pick it up: it moves through a pipeline — an explicit sequence of steps, checks and approvals — and AI agents do the work inside those steps.
The point is not “an AI wrote some code”. The point is that every task goes the same way, the quality gates are the same, and afterwards you can see where the time actually went.
The three things you work with
Section titled “The three things you work with”Projects hold everything else: a task workflow, repositories, pipelines and people. One project usually maps to one product or service.
Tasks are the unit of work — the same statuses, types and fields you would expect from a tracker, plus one addition: a task can be handed to a pipeline.
Pipelines are the interesting part. A pipeline is a graph of steps:
- AI steps — an agent gets a prompt and does the work (plan, implement, review);
- Checks — a command that must pass, like
npm test, or an automated review; - Approvals — a human decides before the run continues;
- Set status — move the task in the tracker as the run progresses;
- Commit — record the produced changes.
Each step declares its outcome, and transitions decide what happens next — including what happens when a step fails or a reviewer sends the work back.
Where the agent runs
Section titled “Where the agent runs”Kodel does not ship its own model or force you into one. The default and simplest setup is your own coding agent on your own machine: you point Claude Code, OpenCode or Codex at Kodel over MCP, and it executes the steps locally while the server owns the graph — which step is next, which gate is open, what has already passed.
That is the flow these docs walk through. It needs no API key inside Kodel and no build infrastructure — your agent, your subscription, your machine.
What you get at the end
Section titled “What you get at the end”Once a task has been through a pipeline you can see, per run: which steps ran, what each one reported, how long they took, what changed in the code, and — on the analytics screen — how that compares across the project: how much time is agent work versus waiting for people, and how often work comes back for rework.
Install the platform → Install.