Reason Agent
Autonomous software engineering agent that can inspect repositories, edit code, run commands and tests, investigate issues, and return tested changes for developer review.
About Reason Agent
Reason Agent is an autonomous software engineering agent from Reason Machines. Reason Machines describes its agent as a coding harness built around a minimal set of general purpose tools. The model determines how to inspect a repository, make changes, and test its work rather than following a fixed sequence of task specific procedures.
The agent can read repository files, execute shell commands, edit source code, run tests and benchmarks, investigate software behavior, and iterate based on results. Official examples include migrating code between programming languages, investigating performance problems, fixing application issues, building software, and returning changes through pull requests for review.
Reason supports local and cloud agent workflows. Features identified as local can operate against files on the developer's device, while cloud sessions run in isolated compute environments. Cloud sessions store messages, workspaces, attachments, and checkpoints so sessions can run and resume. Reason also supports scheduled codebase sessions and connected services including source control providers and MCP servers.
Key Features
- Autonomous Repository Work: The model decides how to inspect a repository, make changes, and test its work using a set of general purpose coding tools.
- Code Editing and Command Execution: Reason can read files, edit source code, execute shell commands, and run tests or benchmarks during an engineering task.
- Local and Cloud Agents: Local features can work against files on a developer's device, while cloud sessions execute inside isolated compute environments.
- Persistent Cloud Sessions: Cloud session messages, workspaces, attachments, and checkpoints are stored so agent sessions can continue and resume.
- Connected Tool Access: Agents can connect to source control providers, MCP servers, communication services, project management services, and other integrations with approved permissions.
- Human Review Workflow: Reason can prepare pull requests that developers review and merge, providing a human decision point before changes enter the main codebase.
Pricing
Contact the sales team for pricing information.
Use Cases
- Investigating and fixing software bugs
- Modifying, migrating, and refactoring code
- Diagnosing performance issues and validating improvements with tests or benchmarks
- Performing codebase research and preparing changes for pull request review
Pros & Cons
Pros:
- Allows the model to independently determine how to investigate, modify, and test a repository rather than requiring a fixed task workflow.
- Supports both local execution and isolated cloud agent sessions that can run and resume.
- Supports connected source control services and MCP servers with permission controlled actions.
Cons:
Integrations
GitHub, GitLab, MCP servers, OpenAI, Anthropic
FAQ
Last edited
October 5, 2026 at 7:53 PM by Venkatraman Chandrasekaran
