Developer hub
Choose your tools. Test them on real work.
Start with the job you need done, compare source-linked facts, and trial a small shortlist in your repository. Benchmarks and editorial scores are useful signals; your acceptance checks decide.
Start with your workflow
Work in a repository
Terminal agents for existing codebases. Check command approval, model access and usage costs.
Compare this category →Choose an AI editor
Compare editor workflows, subscription terms and supported models before migrating a project.
Compare this category →Review pull requests
Compare code review tools, then measure useful findings and false positives on your own changes.
Compare this category →Build an agent
Frameworks for custom workflows. Inspect deployment, model support and orchestration requirements.
Compare this category →Coordinate coding agents
Harnesses and orchestration tools. Check isolation, approvals and how changes are merged.
Compare this category →Run models locally
Compare inference platforms and verify hardware requirements, licenses and network behavior.
Compare this category →A fair trial you can repeat
- Give both tools the same small task, repository commit and acceptance checks.
- Record the tool and model versions, settings, elapsed time, actual cost and any manual fixes.
- Run the existing tests and inspect the diff for correctness, security and unnecessary changes.
- Keep failed attempts in the result. Choose on quality and total effort for your work.
Know what each number means
- Model benchmarks: evaluator results for an exact configuration. Latest board snapshot: 2026-11-12. Missing scores are not zeros.
- Agent board scores: editorial composites using previously generated critic opinions, measured adoption and source coverage. They are not task success rates.
- Source checks: a verification date tells you when a row was reviewed. A source link alone does not mean the information is current.
Use the open dataset
Read the JSON API without an account or API key. Agent records include pricing, model support, source URLs and verification fields. Data is licensed CC BY 4.0; credit AgentBoards when reusing it.
curl -fsSL https://agentboards.org/api/v1/agentsKeep a useful shortlist
Save tools and model comparisons on your device. Share exact model configurations with your team, or download a dated comparison image for a decision record.
For repeat visits: release updates and the RSS feed.
Help keep the reference accurate
Found an outdated price or missing compatibility detail? Submit a correction with its source. Building a tool? Claim the listing and respond on the record.