Direct and check your project's software work in chat.

Chat works in one project at a time. Send a Request exactly as you typed it. Drafts need your Confirm. From chat you can confirm Process Now, pause or resume, retry a blocked task, and refine a done Request. Status and pull request questions answer from project records. A rail shows Running, Queued, Recently landed and Needs you. Export a conversation as JSON, delete it, or delete chat history. Keelen pays for chat inference; coding uses your own engine. Chat is for workspace owners and members on a paid plan or an eligible trial.

Connect Claude Code, OpenAI Codex, GLM, Kimi or Grok. Use your own supported subscription or API key. Grok accepts a SuperGrok sign in or an xAI API key. Pick an engine for each project. Coding-engine use stays on your provider account. If you switch engines, your roadmap and history stay. Engine availability can differ by role. The OpenRouter coding-engine route stays limited to Keelen's own test and benchmark projects. It is off for customer repos. Separately, when hosted chat is on, Keelen pays for that inference through OpenRouter and DeepInfra.

No. Keelen runs on whatever credentials you connect at setup. That can be a Claude subscription login or API key, a Codex login or OpenAI API key, a GLM key, a Kimi key, or a SuperGrok sign in or xAI API key. We don't resell tokens. Coding usage bills to your own provider at cost. Keelen pays for chat. If you do own a Claude Pro or Max plan, Keelen puts it to work. The loop runs on the subscription you already pay for.

Five gates stand between the AI and your main branch. Where a verified project test command supports it, new tests are applied without the implementation first and must fail. Tests born green are rejected. An independent adversarial reviewer reviews the diff. That reviewer shares no context with the dev run. When a verified project test command is available, your test suite runs on a clean checkout. Configured required CI checks protect a merge from a red result. Gated auto-merge has a review window. Manual merge means Keelen opens the PR and you click merge. Branch only mode produces a branch with no PR. Plan review is a separate opt in. When it is on, a human approves the plan before any code is written. When the guards are on and can check the change, they block detected assertion removal or weakening within their supported scope. Supported exceptions can let those changes through. A declared exception does not prove a human approved it. The guards can be turned off. Missing inputs can make them skip the check. They do not catch every way a test can lose meaning. The tests and code still need human review.

Every failed run is sorted into one of 30 failure kinds. Infrastructure blips and provider limits retry on a budget. They never count against your work. A real dead end becomes one plain English card in your Needs-you queue, with a recommended action. The project pauses instead of burning your quota. Iterations are also capped by wall clock time, about an hour for a dev run. Nothing spins forever. You resolve the card, hit resume, and the loop continues.

Keelen supports TypeScript and JavaScript, including Next.js and Vite web apps. Those web apps get preview server QA with a headless browser. It also supports Python, Go, Rust, C/C++, Godot, and Roblox with Rojo/Luau and the Lune test runner. Monorepos are supported, with test targets per package. A repo with no tests at all can still onboard. Keelen runs with an honest no test posture until its own pull requests add a test suite.

Your repo is cloned into an isolated single use VM. That VM is destroyed when the run ends. There is no persistent volume. Code and credentials on disk die with the machine. Some things do persist on our side. We keep iteration logs. Secrets in them are scrubbed automatically. We also keep PR metadata and timeline metrics. Your credentials are encrypted at rest. The keys rotate independently and are deleted the moment you revoke them.

Copilots autocomplete while your hands are on the keyboard. Chat coding agents do one task per prompt, then wait for the next one. Keelen runs a standing process. It holds your roadmap and picks its own next task. It works around the clock. Every change passes through test and review gates. It recovers from flaky CI and merge conflicts on its own. It also collects lessons and steering rules that persist across runs. You direct and check that work in chat. Chat adds no editor and no autocomplete. Coding runs on your own AI subscription, at cost. Keelen pays for chat inference. Chat is for workspace owners and members.

Both. The loop is the product and the graph is the machinery. Keelen is not one model in a while loop. It runs parallel lanes. They cover intake, planning, dev, review and recovery. Scheduler policy routes them over durable Postgres state. Circuit breakers sit on every retry cycle. Human checkpoints sit where you configure them. That covers what graph based agent frameworks promise. It gives you branching, parallelism, checkpointed shared state and a human in the loop. Keelen runs all of that as a crash safe control plane instead of an in memory graph.

Yes, per project. Require plan review, which is a product sign off and/or a developer sign off on every plan before any code is written. Set merge to manual and Keelen opens the PR but leaves the merge button to you. Or choose branch-only and it pushes branches without opening PRs at all. You can also pause one project or your whole fleet at any time, and revoke GitHub or model access in one click.

Yes. Keelen ships a public MCP server at https://keelen.ai/mcp over streamable HTTP. Add it to Claude Code, Claude Desktop, Cursor, Windsurf, the Codex CLI or any MCP client. From chat, 43 tools let you sign up and create projects. They also submit requests, reorder the roadmap, check project status and resolve escalations. Signup and email verification are tools too. So you can connect before you have a key. A request sent over MCP enters the same intake pipeline as one typed into the dashboard.

One iteration is one machine run. That can be a PM planning pass or a dev run. Manual runs and scheduler triggered runs count the same. If you hit your monthly cap, new runs pause until your cycle renews. Or top up instantly with an add-on pack from the billing dashboard.

Yes. The Security scan is a deep whole repository audit. It runs in an isolated VM with a read only single repo token. It works a fixed checklist. It covers code, secrets (including git history), dependencies and infrastructure. You get ranked findings plus a coverage map with no silent gaps. Any finding can be sent to the loop. The loop fixes it and ships a tested pull request. The scan is included on every paid plan.

The Controls scan reviews your repository. It looks for per control evidence. It checks Cyber Essentials v3.3, CMMC Level 1 and CMMC Level 2 (NIST SP 800-171 Rev. 2). The Legal scan maps code observations. It points to commonly cited obligations. Those include GDPR, CCPA/CPRA, COPPA and TCPA. Both are honestly bounded engineering reviews. They are not certifications. They are not legal advice. Both can route code fixable gaps into the loop as tested pull requests.