From a repo to verified live.
PATH is a deployment control plane that runs beside your cluster. Point it at a repo; it works out what the app needs, proposes a plan in plain language, and once you confirm, builds it and gives it an address. Git stays the source of truth. Every change has a name on it.
You're on the list ✓
We'll be in touch before self-serve opens.
- Read the repository · Python, needs Postgres0:04
- Proposed a plan · 1 service, 1 database0:06
- Confirmed by [email protected]0:19
- Built · no Dockerfile needed1:02
- Committed to GitOps · a3f9c1e1:04
- Verified live · billing-api.uat…1:11
Deploys don't fail loudly.
They quietly do nothing.
Shipping one service means wiring several repos together by hand: app, chart, environment, cluster config. Get one line wrong and nothing errors. It just doesn't happen. The GitOps engine underneath is sound. The human interface to it is the gap. PATH is that interface.
Four repos, one change
Every new service is a scavenger hunt across code, charts and environment config.
Silent misses
A typo in a values file doesn't fail. It just never deploys.
Platform team as a queue
Every deploy question becomes a ticket for the three people who understand the stack.
Audit by archaeology
"Who changed prod, and why?" means reading Git history and Slack.
The AI proposes. The platform gates.
The human decides.
PATH was designed for teams that answer to a security reviewer. These aren't settings you switch on; they're how the system is put together.
Git is the only write path
Every change PATH makes is a commit to your GitOps repo, attributed to the person or machine that asked for it. Your existing GitOps engine applies it. The audit trail isn't a feature, it's a by-product.
Read-only by construction
PATH reads your cluster; it never writes to it. The connectors for certificates, load balancers and databases have no write methods at all: a write isn't a permission we withhold, it's code that doesn't exist.
Secrets go in, never come out
Values are written straight to AWS Secrets Manager and shown back only as a reference and the last four characters. No API call or AI tool can return a secret value.
AI drafts, a policy checks, a person confirms
AI reads the repo and drafts the plan. A fixed policy check sits between that draft and anything real, and nothing deploys until someone confirms. Inference runs on Amazon Bedrock through an assumed IAM role, so there's no model API key to leak.
One core, four doors
Web, CLI, API and MCP are equal peers over a single action registry. What your editor's AI assistant can do is exactly what the UI can do, with the same checks.
One platform. Four ways to run it.
Start on our cloud, bring your own when your team grows, or run the whole thing yourself, with a forward-deployed engineer if you want one. Same app model, same commands, same Git-backed audit trail on every plan.
Individual
Invite onlyFor builders shipping their own apps, without a platform team.
Runs on PATH's cloud · usage-based
- Your own workspace, signed in with Google
- Deploy from a GitHub URL or a folder
- A preview for every pull request
- Variables & secrets; secrets are write-only
- Storage with copies, backups and staged restore
- Network-isolated from every other workspace
- Permanent apps that never expire
- Out of balance? Apps stop, never get deleted, and revive on top-up
- Postgres & Redis datastores in one click
- Custom domains with HTTPS
- Logs, live tail & metrics
- Console into a running app, leased and audited
- Roll back to any verified deploy
- Deploy from your editor's AI assistant (MCP)
- Personal deploy token for CI and AI agents
Startup
Coming soonFor teams who want their apps in their own cloud, with guardrails.
Runs on your AWS · your GitHub org
Everything in Individual, plus:
- Connect your cloud: your AWS account, where your apps run, your GitHub org
- Your own deployment repo; every change is a commit you own
- Named environments: uat, prod and more
- Four-eyes approvals for deploys, resources, secret references and rotations
- Playgrounds that expire, and "give it a permanent home" when one sticks
- Members with access levels
- Service principals for CI and AI agents
- Templates: save a deployment plan, reuse it
- Stack view across every environment
- Webhooks & log drains
- Audit log
Enterprise
Coming soonFor regulated organisations that need the control plane inside their walls.
Self-hosted · contract
Everything in Startup, plus:
- Self-hosted control plane: one binary plus Terraform
- Runs entirely in your cloud, next to your apps
- Sign in with your identity provider
- Your own secret store
- Air-gapped installation
- Licence-based, on contract terms
Enterprise Plus
By engagementFor organisations that want it shipped, not just licensed: the platform plus a forward-deployed engineer.
Your cloud · engagement-based
- Security review co-authored from week one
- No lock-in, no long contracts
Everything in Enterprise, plus:
- Ground Truth1–2 weeks embedded. The Field Report: your 3 real problems and the one to ship in 30 days. Credited toward the build.
- Forward-Deployed BuildFixed-scope sprints to production, past Security, documented and handed over.
- Keep-It-AliveOptional retainer to run it and level up your team. Cancel any month.
✓ in the plan◐ rolling out during the invite-only previewNo prices yet: invitees and the waitlist hear first
Private preview.
Opening with a waitlist.
PATH runs today in a private environment. Self-serve sign-up isn't open yet, and the waitlist is how we choose who gets in first, so each team gets onboarded properly rather than left to a docs page.
Questions about PATH.
Does PATH replace our GitOps tooling?
Can I use it from Claude, Cursor or another AI assistant?
Does PATH need write access to our cluster?
What will it cost?
Is PATH part of an FDE engagement?
Join the PATH waitlist.
Tell us what you'd deploy first. We read every reply, and early teams shape what ships.
You're on the list ✓
We'll be in touch before self-serve opens.