An AI agent in every role: one plans, one reads the code, one writes it, one tests, one reviews. They work as a team and bring the work as a pull request. The merge is your call.
The cards below are samples; real work runs in your own repository.
Takes your request or a suggestion from the report, clears up the ambiguity and splits the work into small steps.
Maps the structure, your rules and the domain knowledge, and finds where the change will land.
Follows your architecture and your rules. Every task gets its own branch.
Writes and runs the unit tests for the change; the work does not move on while a test is red.
The change is reviewed, its docs are updated, it is committed and pushed, and a pull request opens. Leave a comment and the revision agent fixes it.
Without signing up, see how the agent team carries a job on a sample project: plan, your approval, code, tests, review and pull request.
This is an illustrative, sped-up example; real jobs run on your repository and take longer depending on its size.
One panel to hand work to agents, watch them live and measure the result. The screens below are the product itself.
The cockpit starts the day with decisions: plans waiting for approval, stuck jobs and open critical findings at a glance.
Every role has its own desk. As a job moves from planning to code, testing and review, its card flies from specialist to specialist; built for big screens and TV mode.
Planning, reading the code, coding, testing, review, git and scenario. Each stage's output and evidence stays on the job page; the plan never turns into code without your approval.
Quick scans and AI-driven smart testing explore your app in a browser and group accessibility, security, form, mobile and console problems by module. One click turns a finding into a fix job.
Merged jobs, cost per job, intake-to-merge time and its bottleneck, first-pass acceptance and engineering hours saved, compared with the previous period.
Screens are shown with sample data.
Not a generic AI: a team that learns your code, your rules and the way you work.
Events from Sentry, GitHub, GitLab, Linear or your own webhook land on the board as cards. Repeats are counted, and a closed error that comes back is caught as a regression.
Your code and documents are read once and split into meaningful pieces. When a job arrives the agent gets only the modules, rules, screens and APIs that matter for it.
Methods like "reproduce a bug with a test first" or "this is how we write migrations" become skills. When the agent finds a reusable method in a job it proposes one; nothing is used until your team approves it.
The screen list comes from the route definitions in your source code and from exploration. Smart testing gives priority to screens nobody has touched.
Agents run on Claude. In your own runner (on-premise or your own server) you choose which account and which cloud region the model is called from. Assign a model per stage, for example a strong one for planning and a fast one for the commit stage.
No new tool to move to; Facimus connects to the places your team already works.
Whatever each person does in a software team, an agent does here. When one finishes, it hands over to the next; comment on the pull request and the revision agent brings back a fix.
What to know before you open your repository to the agent team.
Every job happens on its own branch and arrives as a pull request. Agents are never given the repository key; the git gate in the runner allows pushes only to the job's branch and refuses pushes to protected branches. For extra assurance, keep branch protection on at your git provider.
By default, no code is written before you approve the plan; only you can turn on continuing without approval. The platform never merges a pull request itself.
Agents take work only in the repositories you add. We recommend scoping the access key to those repositories; you can revoke access at any time.
In the cloud, development jobs run in an isolated runner reserved for each workspace, and it shuts down when idle. Workspace data is separated at the database level.
Your access keys are stored encrypted and handed only to the runner doing the job, at the moment it runs.
The plan, the change, test results, cost and who approved what stay in the job's history.
From product teams to enterprise IT: the same agent team, with your rules.
No payment or card details at sign-up. Your account starts on the Starter plan; when you grow, you request a quote from the panel.
To try it with one repository and a small team.
New accounts start here
For teams with several repositories and a steady flow of work.
Let's talk pricing
Enterprise sign-in, user provisioning and every deployment option.
Let's talk pricing
The AI quota is what agents may spend in the cloud in a month; on-prem and on your own server, AI costs are yours.
No. Every job is done on its own branch and arrives as a pull request; merging is always your call. By default no code is written before you approve the plan either.
Either in the cloud, in an isolated environment reserved for your workspace, or in a runner inside your own network. With an on-premise install the whole platform runs in your data center.
Agents run on Claude. In the cloud the platform's model account is used; in your own runner you choose the Anthropic API, AWS Bedrock, Google Vertex AI, Microsoft Foundry or your Anthropic-compatible corporate gateway. A different model can be assigned per stage.
On GitHub and Bitbucket it opens the pull request itself and follows reviews. On other git servers it pushes the branch and you open the pull request.
Set a monthly budget cap and a per-job cap; when the cap is reached no new job starts. The Value dashboard shows cost per job and return.
Right after signing up you can watch the agent team work on a sample project. The first job on your own repository starts once the repository is connected and the runner is ready.
Your code is already in your repository. Jobs, rules, skills, scenarios and findings are exported as a single file; the audit log is downloaded separately as CSV or JSON.
Early access opens soon. Once it does, you can start right here.