Ottili Coder

Ottili Coder is the chat-first software-automation product for Ottili ONE. Describe a software mission in plain language and Coder turns it into task queues, runs dynamic agents across local, cloud, and hybrid execution, automates Git, validates every change, and deploys — all authenticated through your Ottili account.

What it does

• Chat-first mission input*: describe a software mission in plain language — goals, repo context, constraints, and deployment target. Coder turns it into a structured, reviewable task queue instead of a dead-end prompt.
• Dynamic agents*: role-based agents (backend, frontend, infra, QA, security) run tasks locally, in the cloud, or hybrid, with logs, ownership, and validation visible.
• Validation loops*: every task runs through type checks, tests, and security gates with retries and explicit failure states before it is marked done.
• Automated Git*: Coder opens branches, commits, and prepares pull requests from agent output — reviewable, attributable, and reversible.
• Deployments*: passing runs ship to Docker-first targets with deployment status, rollback, and audit kept on the same run record.

Execution modes

Coder runs the same task queue in three modes. Orchestration, logs, validation, and the audit trail are identical regardless of where the work happens — only the runner changes.

• Local execution (Live)*: code and agents run on your own machine through the Coder CLI or desktop app. Source stays in your environment; only the run metadata needed for identity, credits, and audit leaves it.
• Cloud execution (Beta)*: Ottili provisions a managed remote runner. Your repository and task context are sent to a cloud workspace, and results return to the same run record. Cloud runs are in beta: capacity is limited and credit-metered, behavior may change, and it is not intended for production secrets without review.
• Hybrid execution (Beta)*: plan and run lightweight tasks locally, then hand long-running or compute-heavy work to a cloud runner — all on the same run, with one coherent queue. Hybrid depends on cloud runs, so it carries the same beta maturity.

Runner requirements

• Git, with read/write access to the repository you want Coder to work in.
• A runtime for the work: the Coder desktop app or a container runtime such as Docker.
• Node.js and Python for the runtimes Coder drives during a run.
• An Ottili account with Coder enabled — this supplies identity, credits, and the run queue.
• Network access to Ottili for authentication, queue sync, credit accounting, and audit.
• Optional: MCP servers and provider credentials for the specific task.

The ottili-coder doctor command checks these before a run starts.

How data flows through a run

1. Mission input becomes a durable run record, not a disposable chat message.

2. Planning expands the mission into a reviewable task queue (recon, decomposition, risk).

3. Execution runs agents in isolated workspaces — locally, in the cloud, or hybrid.

4. Validation gates each task: type checks, the test suite, and security checks must pass.

5. Artifacts are produced through your Git: branches, commits, and pull requests.

6. The record closes with logs, approvals, status, and the audit trail kept on the same run.

Permissions and your data

• Authenticated through your Ottili identity ("Sign in with Ottili").
• Access follows your company and role model; approval gates hold risky or external steps.
• Local runs keep source on your machine; cloud runs send repo context to a managed workspace.
• An audit trail (who, what, when) stays on the run record in every mode.

Offline limitations

• Coder is not fully offline: the local runner still needs Ottili connectivity for auth, credits, queue sync, and audit.
• Cloud runs additionally require network access to the cloud workspace.
• Secrets are resolved through Ottili at execution time; they are not stored in plain text on the runner.

Cloud availability

• Cloud runs* — Beta: limited capacity, credit-metered, subject to change.
• Cloud deployment* — Planned: shipping a run result to Ottili Cloud infrastructure is planned and is not yet a public destination. It is not presented as available.
• Current Coder service status (Web, Cloud Runtime, CLI Services, GitHub Integration) is published on the public status page.

Workflow

1. Start a session from the Coder surface or via "Sign in with Ottili".

2. Describe the objective in plain language.

3. Review the proposed changes and diffs.

4. Approve and let Coder apply or open a PR.

Auth

Ottili Coder uses your Ottili identity ("Sign in with Ottili"), so access and audit follow the same company and permission model as the rest of the platform.

Guides

Start here, then go deeper with the focused Coder guides:

• [Installing Ottili Coder](/docs/coder-installation) — CLI and desktop app setup, plus the doctor check.
• [Coder projects](/docs/coder-projects) — how a project scopes repo, docs, and constraints.
• [Coder missions](/docs/coder-missions) — turning a plain-language objective into a run.
• [Coder modes](/docs/coder-modes) — build vs plan mode.
• [Coder task queues](/docs/coder-task-queues) — recon, decomposition, and risk.
• [Coder agents](/docs/coder-agents) — specialized roles and ownership.
• [Local, cloud, and hybrid runs](/docs/coder-local-cloud-hybrid-runs) — runners and availability.
• [Coder and Git](/docs/coder-git) — branches, commits, and pull requests.
• [Coder validation](/docs/coder-validation) — type checks, tests, and security gates.
• [Coder approvals](/docs/coder-approvals) — human-in-the-loop gates.
• [Coder deployments](/docs/coder-deployments) — how results ship today and cloud status.
• [Coder plugins](/docs/coder-plugins) — extending Coder with custom tools.
• [Coder models](/docs/coder-models) — model routing by role and task.
• [Coder troubleshooting](/docs/coder-troubleshooting) — auth, connectivity, and environment.
• [Coder security](/docs/coder-security) — identity, secrets, isolation, and audit.

Tips

• Keep objectives small and verified; let Coder run tests before claiming done.
• Use cloud runs for long-running, well-scoped work and review the resulting PR.