Introduction
Kadmo is an agentic engineering platform that runs AI coding agents as a coordinated team inside your own tools, cloud and models. Instead of one assistant taking an entire ticket, Kadmo assigns defined roles, playbooks and approval gates, so work moves from backlog to pull request with a written record of every step.
What is Kadmo?
Kadmo is a platform for running AI coding agents the way a software team runs: with roles, rules and review points. Engineers, a product manager, a reviewer, a QA agent and other specialists each get their own machine, their own scoped credentials and one narrow job. A playbook choreographs the handoffs, and a gate sits at every handoff, so nothing moves on until it passes.
The problem it addresses is familiar. A single generalist assistant handling a whole ticket tends to produce large, hard-to-review changes, and teams cannot easily see what happened or why. Kadmo splits the work into smaller steps, keeps each context window small and sharp, and writes a comment into the ticket at every step explaining what the agent did and why.
Because agents run inside the tools a team already uses — Jira or Linear, GitHub or GitLab, Slack — there is nothing to rip out and nothing to migrate. Kadmo can be deployed on your own cloud account or your own machine, which keeps source code inside your perimeter. That combination makes it relevant for teams that want AI agents for software development but cannot send code to a third-party service.
Kadmo states that teams see up to 10× engineering output and 24/7 coverage, alongside a 92.5% first-pass acceptance rate; these are vendor figures, not independently verified results.
Key Features of Kadmo
Role-based agents with scoped tools
Kadmo assigns each agent a role — engineer, product manager, reviewer, QA, documentation, security — so it works only with that role's tools and credentials. Permissions stay tight and output stays predictable.
Playbooks with a gate at every handoff
Define a playbook once and Kadmo runs the sequence, pausing at gates you control. Every step posts its reasoning into the ticket instead of leaving reviewers to guess.
Works inside your existing toolchain
Kadmo connects your task board, repository and cloud, so agents pick up tickets where they already live. Supported integrations include Jira, Linear, GitHub, GitLab, Slack, AWS, Google Cloud and Hetzner.
Sovereign deployment
Run the fleet on Kadmo infrastructure in Frankfurt, in your own cloud account, or on your own machine. No inbound port opens, and each agent works in its own sandbox with scoped credentials.
Governance and audit built in
Set approval policies, spend ceilings per account and workspace, and data residency rules. Kadmo keeps a full, replayable transcript of every run plus an exportable audit log, with SAML and SCIM support for sign-in.
Self-learning skills
A skill discovery agent reads the codebase and past runs, writes down what other agents need to know, and loads those notes into later runs. The second ticket starts smarter than the first.
Ready-made agentic solutions
Kadmo ships with solutions that add value on day one: feature implementation, goal-to-tickets planning, release checks, documentation updates, website builds and a nightly security scan in preview.
Use Cases for Kadmo
Ticket-to-pull-request delivery
Hand a ticket to a software engineer agent; it works on its own branch, opens a pull request and fixes what review sends back. Reviewers stay in the loop and spend less time on mechanical work.
Release and regression testing
A QA agent runs your test suite on the live environment after each change and files failures as tickets. Fixes flow back to the engineer agent and are retested before the ticket closes.
Planning goals into tickets
A product manager agent turns a goal into an epic and splits it into tickets sized for one agent each, then checks results against the original intent.
Documentation that follows the code
A documentation agent reads what shipped, updates the docs and drafts the customer update for approval, so documentation stops lagging behind releases.
Compliance evidence collection
Because every run keeps its transcript, teams can collect evidence as work happens and hand it to an auditor without a last-minute scramble.
How to Use Kadmo
- Connect your tools. Point Kadmo at your task board (Jira or Linear), your repository (GitHub or GitLab) and the cloud account where agents should run.
- Choose roles and models. Select which AI coding agents and models to use — Claude Code or Codex, on Anthropic, OpenAI or Google models, directly or through your own gateway.
- Set the rules once. Define playbooks, approval gates, budgets and residency policies so every run follows the same guardrails.
- Drop a ticket. Assign one backlog item to the fleet and let each agent comment progress into the ticket.
- Review at the gates. Approve or send work back; Kadmo records the decision and the next run picks it up.
Target Audience for Kadmo
- Engineering teams already using AI coding agents who want structure and review points
- CTOs and platform leads with strict data residency or compliance requirements
- Product managers who need goals converted into well-sized tickets
- Agencies and product studios handling many small tickets in parallel
- Regulated organisations that need an audit trail for automated work
Is Kadmo Free?
Pricing for Kadmo is quote-based rather than a published monthly list. The reference material names Business and Enterprise plans, with governance features such as scoped roles, approval policies, budgets, data residency enforcement, run transcripts and SSO included at those tiers. Runs are paid through prepaid Kadmo credit, or through your own model keys where a plan includes them.
| Plan | Price | Highlights |
|---|---|---|
| Business | Custom quote | Agent fleet, roles and playbooks, governance controls |
| Enterprise | Custom quote | Governance plus SAML SSO, SCIM provisioning, residency policies, audit log export |
| Pilot | Start on one backlog item | Set up with you, live within days |
Check the official pricing page for current figures. A short pilot on a single ticket is the simplest way to judge value before a wider rollout.
Kadmo's Pros and Cons
| Aspect | Pros | Cons |
|---|---|---|
| Deployment | Runs in your cloud or on your machine; code stays inside your perimeter | Setup involves infrastructure choices |
| Workflow fit | Works inside Jira, Linear, GitHub, GitLab and Slack | Needs an existing task board and repository |
| Oversight | Role-scoped agents, approval gates, replayable transcripts | Gates add steps some teams may find slow at first |
| Cost control | Per-run token spend reported in the ticket, budgets with ceilings | Quote-based pricing means no public rate card |
| Flexibility | No model lock-in; switch providers or use your own gateway | Best results need clear tickets and playbooks |
Frequently Asked Questions about Kadmo
What do I need to start with Kadmo?
A repository, a task board and one item from your backlog. Kadmo works inside Jira or Linear and GitHub or GitLab, so nothing is replaced, and setup is done together with your team.
Where do the agents run, and does my code leave my perimeter?
You can run the fleet on Kadmo infrastructure in Frankfurt, in your own cloud account, or on a machine of your own. With the latter two, code stays inside your perimeter, and every agent gets an isolated machine with scoped credentials.
Which AI coding agents and models does Kadmo support?
Claude Code and Codex are supported today, running on Anthropic, OpenAI and Google models — direct, through Amazon Bedrock or Vertex AI, or behind your own gateway. Nothing locks you to one provider.
How much does a ticket cost on Kadmo?
Each run reports its real token spend inside the ticket, so a shipped ticket has a visible price, and spend limits can be set per account and workspace.
Is Kadmo free to try?
There is no published free plan. The platform is set up with a team and piloted on one backlog item before a wider rollout, with pricing available on request.
How is Kadmo different from a single AI coding assistant?
A single assistant usually handles one large ticket in one context. Kadmo splits work across specialist agents with small context windows, adds a gate at every handoff, and documents each step in the ticket.
Kadmo Tags
Kadmo, Kadmo review, agentic engineering platform, AI coding agents, AI engineering team, sovereign AI platform, AI agents for software development, self-hosted AI agents, AI code review automation, playbook-driven AI agents, Claude Code and Codex, AI agent governance
