Run your own factory.
We run a software factory every day, for our own products and our clients’. We teach your engineers to run one in your repositories: what to build first, what to check, and where to put people. Most teams start with a workshop or an audit.
| Workshops | Factory audit | Factory build | |
|---|---|---|---|
| Scope | 1–2 days, in your code | 1–2 weeks, fixed fee | Fixed scope |
| For | Teams that use coding agents one at a time and want to run many of them on the same code, safely. | Teams whose agents stall or break things, and teams about to give agents to every engineer. | Teams that know what is missing, from an audit or from their own list. |
| You leave with | A working gate and a team that has run many agents at once | Written findings and the first commits | The missing parts, built in your repositories |
Workshops
1–2 days, in your code
Your engineers build and run a factory on one of your own repositories, with us in the room.
Who it’s for
Teams that use coding agents one at a time and want to run many of them on the same code, safely.
What happens
- Pick one repository and a few real issues to work on.
- Write the guides together: an AGENTS.md, a handful of rules, a first skill.
- Build the first check: a pull-request gate that runs the same command on a laptop and in CI.
- Run several agents at once, each in its own worktree, and watch what the gate catches.
- Move one credential out of reach: the agent calls a tool that holds the key.
What you walk away with
- Guides and a working gate, committed to your repository
- Engineers who have run many agents on one codebase
- A written list of what to build next
Factory audit
1–2 weeks, fixed fee
We read your repositories, pipelines and access model, and find where agents will stall, where they will break things, and where they could already run.
Who it’s for
Teams whose agents stall or break things, and teams about to give agents to every engineer.
What happens
- An hour on a call about what you ship, who ships it, and where it hurts.
- We read your repositories, pipelines and access model against the five parts of a factory.
- You get written findings, ranked by cost and payoff.
- We make the first commits: an AGENTS.md, a PR gate, or whatever unblocks you first.
What you walk away with
- A readiness assessment across all five parts
- Written findings you can hand to your team
- The first commits, in your repositories
Factory build
Fixed scope
We build the parts of a factory your organization is missing: guides, checks, access and infrastructure, against a scope agreed up front, and teach your team to run it.
Who it’s for
Teams that know what is missing, from an audit or from their own list.
What happens
- A written scope and a definition of done, agreed before work starts.
- The work, as commits and pull requests in your repositories, through your review.
- A handover your team can run without us.
What you walk away with
- The missing parts of the factory, built and running
- Everything in your repositories, owned by you
- A team that knows how it works
What we build into your factory
The five parts a factory needs. An audit reads your repositories against them; a build fills in the ones that are missing. How our factory runs
-
A codebase agents can read
Guides
Layout on disk, AGENTS.md and CLAUDE.md, rules, skills and specs: what an agent reads before it writes a line.
- Repository layout and module boundaries agents can follow
- AGENTS.md and CLAUDE.md operating manuals
- Rules, skills and spec templates for your recurring work
-
Many agents, one codebase
Orchestration
Parallel agents in isolated worktrees, agent hosts, MCP servers, skills, triggers and scheduled jobs: dozens to hundreds of agents on one repo.
- Agent hosts and per-worktree environments
- Triggers, scheduled jobs and hand-off conventions
- Module boundaries that let parallel changes merge cleanly
-
Access without exposure
Capabilities, not credentials
Agents get capabilities, never credentials: secrets brokered by a vault, MCP servers scoped to the job, least privilege, and an audit log.
- Secrets brokered by a vault, out of every context window
- Scoped MCP servers and per-agent, per-host keys
- Least-privilege defaults and an audit trail
-
Verification at volume
Sensors
Tests, QA pipelines, holdout scenarios, review agents and merge gates that can check hundreds or thousands of changes a week.
- A PR gate that runs locally and in CI with the same command
- QA pipelines and holdout scenarios agents cannot game
- Review agents and merge rules sized for agent throughput
-
Reliability at agent speed
Infrastructure
Progressive delivery, rollback and observability, so production at thousands of changes a week is as steady as it was with weekly deploys.
- Deploys that verify themselves and roll back in one step
- Progressive delivery and environment parity
- Observability agents and people can both read
Start with the writing
Free, and specific enough to check.
Start with a workshop or an audit.
Tell us what you ship, how many people and agents work on it, and what keeps breaking. Talk to our agent now, or email us and an engineer answers.
A voice call with our AI agent. Email us instead