Harmont
Website unreachable since Nov 2, 2026
Harmont is a continuous integration platform that runs tests directly on a developer's working tree, enabling uncommitted code to be validated before it is pushed. Pipelines are written in real Python or TypeScript code instead of YAML, and the system outputs machine-readable results with MCP support so AI agents can read failures, fix pipelines, and re-run. The tool combines persistent state with stateless scaling, allowing parallel forks and CI completion in seconds.
- Developer Tools
- Software Only
- Artificial Intelligence
Funding
Founders
Founder details are not available yet.
Product
Problem
Traditional CI/CD platforms test code only after it is committed and pushed, meaning broken builds reach the team and shared repository before failures are caught. Pipelines are typically written in YAML, which is difficult to type-check, test, and reuse, and the results are not structured for automated AI agents to interpret and act upon.
Solution
Harmont provides a continuous integration platform that runs directly on a developer's working tree, allowing uncommitted code to be tested before it is pushed. Pipelines are written in real Python or TypeScript code rather than YAML, making them typed, testable, and reusable. The system outputs machine-readable results and supports MCP, enabling AI agents to automatically detect failures, fix pipelines, and re-run. Harmont offers persistent state with stateless scaling, using warm disks and avoiding artifact shuffling, so forks run in parallel and CI completes in seconds. A single static binary is all that is required to install the tool, with no additional dependencies.
Target Audience
Primary users are software developers and engineering teams who want faster, more reliable CI feedback before code is pushed, as well as AI agent developers and teams integrating automated pipeline fixing and re-running into their workflows.
Features
- Runs CI on the working tree to test uncommitted code before push, preventing broken builds from reaching the team
- Pipelines written in real Python or TypeScript code instead of YAML, providing type safety, testability, and reusability
- Machine-readable output with MCP support so AI agents can read failures, fix pipelines, and re-run automatically
- Persistent state with stateless scaling, using warm disks and parallel forks to achieve CI completion in seconds
- Single static binary installation with no additional dependencies
- Toolchain integration with `uv` and `cargo` to bring lint, typecheck, and test steps into pipelines