Kodel is the management layer over Claude Code, OpenCode, and your agents. Describe workflows as code, run them through guardrails, and see every step — not a black box, a governed engineering process.
Pre-launch · SaaS-first · Enterprise self-hosted available
Agent → check → human gate. The server owns the graph — agents can't skip a step or fake an outcome.
See where every task is, which checks passed, who approved — pipeline-level visibility, not terminal logs.
Your team hired coding agents for speed. The process stayed ad-hoc — and predictability dropped below expectations.
Unclear what was done, by whom, and through what process — prompt in, diff out.
Can't verify that tests, lint, or human review actually happened before merge.
Each developer has a different prompt and level of control — no team standard.
No audit trail, no reproducibility — you can't debug a failure or repeat a success.
If a third of your team already uses coding agents regularly and at least one of these happened in the last three months — you're in the right place.
The agent faked a result, tests failed, or nobody can explain what happened on a task.
A third of the team uses agents, but the process is still "how Ivan does it" — not a standard.
CEO, board, or compliance asks: "How is AI-assisted development governed in your engineering org?"
A new engineer can't repeat a colleague's successful agent run — knowledge lives in DMs, not in process.
You need to prove that AI-assisted changes go through a controlled, documented process — not trust in a chat log.
Your developers already use coding agents. Kodel helps you turn ad-hoc prompts into a repeatable, auditable engineering workflow — without slowing the team down.
Not another agent — an operational model for AI-assisted development you can audit, standardize, and embed in engineering culture.
Replace "everyone has their own prompt" with runbooks your squad can apply on day one. See where tasks block and where human review is required.
Config-as-code pipelines, MCP integration, and observability hooks — SaaS by default, self-hosted when data must stay in your perimeter.
kodelctl applyghcr.io/kodelai/* — secrets stay yoursCTO? Subscribe and run one pipeline end-to-end — get an audit trail on a real task without deploying infrastructure.
Claude Code and OpenCode stay the executors. Kodel defines the process, enforces gates, and holds the source of truth for every task.
Projects, workflows, skills, and pipelines in YAML — applied with
kodelctl apply. A DAG of agent, check, and human
nodes your whole team shares.
Instructions injected per step — intent fixed in code, not reinvented per developer.
Automated commands and tests with explicit pass/fail — the server records the result.
Approval points where the process must stop for a person — enforced, not optional.
The server owns the graph, gates, and final status. Agents report one step at a time via MCP — you see the full path, not just the final diff.
Where the task is now, which steps completed, what blocks progress.
Which automated checks ran, who approved at human gates — audit-ready by design.
Bottlenecks, time on human gates, recurring failures — Datadog for coding agents.
CI/CD gave observability for deploys. Kodel does the same for the era of coding agents — not a replacement for your agents, a control layer above them.
Every task follows a pipeline you define. The agent executes one step; the server decides when it's done.
tier-0 / local — agent on the developer's machine · tier-1 — server-managed dev-box
Describe your project, tracker, and workflow in YAML. Pick a runbook template or write your own pipeline — version it in git.
Your coding agent connects via MCP, runs one step at a time, and reports back. It cannot skip ahead or claim success without passing gates.
Every node is an observable event — skills in YAML, check exit codes, gate status on the server. Full trace per task and per team.
Human gates where review is mandatory. Tracker Bridge keeps Jira or Linear in sync — one ticket, one traceable AI workflow.
Sign up for SaaS and run your first pipeline end-to-end — no infrastructure to deploy.
Tech Lead? Describe one real team workflow in YAML in 30 minutes and compare it to how your squad works today.
Core capabilities shipping in the first release — productized workflows on top of the Kodel pipeline engine.
Ready-made YAML pipelines for bugfix, feature, refactor, dependency updates, and incident fix. Your team's playbook — applicable in a day, not invented from scratch.
Dashboard for CTOs and Tech Leads: tasks by pipeline stage, time on human gates, recurring check failures, and who approved what — team-wide, not per terminal.
CI/CD checks the artifact after merge. Agent CI controls the process before the PR — lint, tests, security scan, and review gates enforced on every AI task.
AI does the first pass against your checklist; human gates guarantee the final word on sensitive paths. Review by process, not by trust in a chat log.
Jira, Linear, or GitHub Issues → Kodel pipeline. Ticket status reflects pipeline stage — one ticket, one traceable workflow, no shadow process.
Tier-1 server-managed sandbox: agents and checks run in isolation. AI-assisted development inside a perimeter that passes security review.
Every step, human gate, and check in an immutable audit log tied to the task and approver. AI in development with a provable process — not trust in a chat.
Illustrative view of the Observability Console — task flow, gates, and tracker sync in one place.
Start on managed cloud in minutes. Move to self-hosted or Secure Dev Box when compliance requires it.
Default deployment — fast onboarding, no infrastructure to operate. Enterprise-grade security for most teams out of the box.
docker compose up in your infrastructure — images from
ghcr.io/kodelai/*. Data, secrets, and audit trails stay
in your perimeter. Start with
kodel-compose.
Role-based access for pipelines, dashboards, and configuration. Every step, gate, and approval logged — server-owned, not agent-reported.
Agent-agnostic. Works with Claude Code, OpenCode, and other MCP-compatible coding agents — your tools, Kodel's process.
Pre-launch pricing is tailored to team size. Subscribe for details when pilots open — tiers below reflect product scope, not final pricing.
Fast entry for squads adopting coding agents with guardrails.
For mature teams scaling AI-assisted development across projects.
Self-hosted deployment for regulated and security-first organizations.
Quick answers before you subscribe — we'll share more detail as features ship.
Cursor and Claude Code help developers write code. Kodel is the control layer above them — it defines the pipeline (agent → check → human gate), enforces guardrails, and gives leaders observability across the whole team. Your agents stay; the process becomes repeatable.
CI checks the code after merge. Kodel orchestrates the process while the agent works — before and during the task — with gates and human approval. Agent CI ensures no AI task closes without the checks your team requires.
SaaS is the default — sign up and run pipelines without deploying infrastructure. Enterprise
customers can run Kodel self-hosted with Docker Compose and
ghcr.io/kodelai/* images when data must not leave the
perimeter. See kodel-compose
to evaluate locally.
The coding agent stays the same — Claude Code, OpenCode, or whatever your team already uses. Kodel only defines the process once in YAML so nobody reinvents it per task. Less coordination in Slack, not another IDE.
For most teams, managed SaaS with enterprise-grade security is enough. If code and audit trails cannot leave your perimeter, Enterprise self-hosted runs entirely in your infrastructure — you hold the secrets and the data.
The earlier you standardize, the less tech debt you accumulate from ad-hoc agent usage. Teams that wait usually fix process chaos after the first incident or board question — not before.
Direct onboarding with the founders, influence on roadmap priorities, and early access to Runbooks, Observability Console, and Agent CI. Typical pilots start with one runbook (e.g. bugfix) and expand from there.
On SaaS, most teams target a first pipeline run within the first days. Apply a runbook template, connect your agent via MCP, and trace a real task end-to-end.
You'll get occasional product news: what we shipped, SaaS and pilot openings, and early access invites. No spam — unsubscribe anytime.
Kodel is in active development. Pick the next step that matches your role.
Subscribe for SaaS early access, run one pipeline end-to-end, and get an audit trail — no infrastructure to deploy.
Describe one real bugfix or feature flow in YAML and compare it to how your squad works today — runbooks included.
Try self-hosted via kodel-compose and a local pipeline example — data and audit trail stay inside your infrastructure.