Gitmatic

For small businesses

A ledger for what your agents change.

When an agent or a person changes your infrastructure or code, Gitmatic records what changed, who or what did it, why, and how to undo it. You can roll a bad change back, answer an auditor or a customer's security questions, and keep agents from making unapproved changes. The record also plugs into the compliance tools you already use, which makes audits easier.

The problem

What you deal with today, and what you actually need.

“What did the agent change, and why?”Scroll chat transcripts, kubectl history, CloudTrail.One timeline: intent, diff, live result, agent, and human.
An auditor asks how you control AI changes.Write a policy doc nobody enforces.A signed, hash-chained record you can export, or send into the compliance tools you already use.
A bad agent change lands in production.Find the old YAML and re-apply it by hand.One-click rollback of the intent, recorded as a new operation.
Every agent shares one kubeconfig.Shared admin credentials.Per-agent identity, tied to a human, scoped by ledger.

How it works

A clear record of every change, without handing over the keys to your cluster.

Gitmatic doesn't host your code, and it never holds cluster credentials. The record is written to a ledger you control. The cluster agent talks outbound only.

01

See why it happened

Every change carries a declared intent: the reason it was made, the agent, and the human it acted for. An MCP server is optional. A required annotation, or this API, is enough.

02

Keep a record you can trust

Each operation is appended to a hash-chained log: type, target, diff, policy result, observed state, and the previous hash. History is never rewritten.

03

Roll back as a new entry

Rollback and rollforward are new operations that point at an earlier one. You can verify the chain at any time. The in-cluster agent is what actually rolls production back.

The record

If someone edits a record after the fact, you'll know.

Chain intact

Each operation's hash covers its payload and the previous hash. If someone edits a record after the fact, verification fails and the gauge moves off the intact mark.

The timeline

A ledger, not a chat log.

The timeline is meant to be easy to scan: sequence, intent, type, policy decision, and the content hash. Code changes and control-plane changes share one log, including mixed changes that touch both.

You install the agent in your Kubernetes cluster with one Helm command and a token.

Integrations

Plugs into the compliance tools you already use.

Your change record and audit evidence flow into Drata and other compliance tools, so audits are easier.

Drata

GitHub

Slack

Pricing

Simple monthly plans, billed through Stripe.

Billing is handled securely through Stripe. It's monthly, and you can cancel anytime.

Agent

Free

Apache-2.0 agent and Helm chart. Needs a Gitmatic account.

Cloud Free

Free

1 cluster, 5 agents, 30-day timeline, email alerts.

Team

$249/month

3 clusters, unlimited agents, approvals, rollback, 1-year retention, evidence export.

Growth

$699/month

10 clusters, SSO, policy packs, 3-year retention, GRC push.