01 / COST
Spend the right model on the right work.
Configure implementation and review models independently. Use lower-cost models where they pass your checks, and stronger reasoning where it earns its cost.
AI CO-WORKERS, IN YOUR WORKSPACE
Bring AI into the places where work happens.
Curie brings AI agents into the workspaces your team already uses. Start with the Software Factory: turn selected engineering issues into reviewable pull requests while your team keeps the final say.
# release-operations •••
✦ Illustrative co-worker conversation
ONE PLATFORM. MANY WORKSPACES.
GitHub · GitLab · Jira · Slack · Bitbucket
START HERE
Meet your Software Factory co-worker.
Put selected GitHub issues through a controlled workflow that plans, implements, checks, and returns a pull request for engineers to review.
Each issue runs in an isolated sandbox. Choose models for implementation and review, apply your approval policy before publication, and keep CI and engineers in charge of what merges.
SOFTWARE FACTORY WORKFLOW / 01
01
ISSUE
Pick a bounded taskGive the co-worker clear acceptance criteria in the issue tracker.
GitHub issue ↗
02
DELEGATE
Curie works in the backgroundFollow planning, implementation, review, and verification as work progresses.
03
REVIEW
Your team accepts the resultInspect the pull request, checks, and cost before deciding what ships.
Pull request ready ✓
THE SHIFT
Agents are getting better. Their workspace is too small.
Most agents sit in a separate app. Your team works across issues, code, chat, and approvals. Curie connects the agent to that shared context, so work can move forward where the team already coordinates.
WHERE AGENTS OFTEN LIVE
IDE · Chat app · Agent workspace
Useful work, isolated from the team’s daily flow.
WHERE THE TEAM WORKS
Issue → Code → Thread → Approval
A co-worker with an identity and a place in the workflow.
01
Like a teammate
Delegate work, ask questions, and follow progress from the spaces you already use.
02
Visible to the team
Work and decisions stay in shared context instead of disappearing into a private session.
03
Under your control
Set model choices, approval boundaries, and budgets for the work you hand off.
WHY DEPLOY THE FACTORY
More reviewable work. Clearer economics.
The factory is designed to help engineering teams move a bounded issue queue while keeping cost and acceptance visible.
02 / BACKLOG
Move selected issues into review.
Admit well-scoped issues with a GitHub label. Isolated jobs can progress in parallel and return a pull request or a clear reason they could not finish.
03 / ROI
Measure outcomes, not just tokens.
Compare accepted PRs, CI results, review effort, cycle time, and total model plus infrastructure cost. Engineers still decide what merges.
WORK WITH THE AGENTS YOU USE
Keep coding locally. Send queued work to the factory.
Use an IDE agent, Claude Code, or Gemini CLI for interactive exploration, issue scoping, and PR review. The factory picks up selected GitHub issues for independent execution. The shared handoff is the issue and the resulting pull request.
Explore the open-source factory ↗01
Work locally with your preferred coding agent to investigate and define a bounded issue with acceptance criteria.
02
Apply the factory label to put that issue in the queue. Curie runs the job in an isolated sandbox.
03
Review the linked PR, checks, and cost in your normal GitHub workflow. Your team controls publication and merge.
THE PLATFORM
One co-worker platform. Built for real work.
The Software Factory is the first use case. The platform gives co-workers a shared foundation to run in your infrastructure and operate across workspaces.
01 / WORKSPACES
Issue · Code · Chat
Native to the workflow
Connect to the tools where work starts and where the result needs to land.
02 / CONTEXT
Slack ↔ Code host
Context that travels
Shared memory facilities help the co-worker carry useful context across workspaces.
03 / CONTROL
Planning · Frontier model
→Execution · Model of choice
Spend with intent
Configure models and inspect usage, with approvals and cost controls close to the work.
OWN YOUR CO-WORKER
Open source. On your infrastructure.
Curie is Apache 2.0 licensed and self-hostable. Start with the Software Factory, inspect how it works, and build from there.
✓ Apache 2.0 ✓ Self-hosted ✓ Your model configuration
Explore Curie on GitHub ↗WORK WITH CURIE
Start with the platform. Go further together.
Adopt the open-source platform on your own, or work with CurieTech on a production rollout and a broader AI transformation program.
01
Explore
Try the open-source platform and prove out a bounded workflow.
02
Deploy
Get support for a production setup and the controls your team needs.
03
Expand
Bring more agents and workflows into the places your teams work.
THE FUTURE OF WORK, TOGETHER
Put your next AI co-worker where your team is.
See the Software Factory in action or talk through a co-worker for your team.