Launch catalog · 5 published FitScores · No pay-to-rank
Tell us the job.We'll tell you the tool.
FitForTool ranks AI and agent tools by the job you actually need done. An agent runs a published test. A person reviews it. The result is a 0–100 FitScore. Nobody can pay for a higher place.
The first published ranking is provisional: coding agents, one TypeScript project, tested 5 Nov 2026. Claude Code 94, OpenCode 93, Codex CLI 90, Cline 88, Google Antigravity 79. Every other candidate is listed as not yet tested, with no score.
- Jobs
- 15
- Candidates
- 99
- Published scores
- 5
- Paid ranks
- 0
Pick the job
- 01Top FitScore 94
Coding agents
Which coding agent should I use to fix bugs and add features in my repo?
11 candidates · 5 tested
- 02Not yet tested
App builders / vibe coding
Fastest way to build a working web app from a prompt?
6 candidates
- 03Not yet tested
AI code review
Which tool reviews my PRs and catches real bugs?
6 candidates
- 04Not yet tested
MCP servers
Which MCP server should I plug into my agent?
7 candidates
- 05Not yet tested
Agent frameworks
Which framework should I build my agent on?
7 candidates
- 06Not yet tested
Browser agents
Which AI can use a website for me?
6 candidates
- 07Not yet tested
Voice / TTS
Best AI voice for narration or a voice agent?
6 candidates
- 08Not yet tested
Video generation
Best short video clip from a prompt?
7 candidates
- 09Not yet tested
Image generation & editing
Which image AI for thumbnails, product shots or game art?
7 candidates
- 10Not yet tested
Research assistants
Reliable, cited answer to a research question?
6 candidates
- 11Not yet tested
Meeting notes
Which notetaker gets action items right?
6 candidates
- 12Not yet tested
Writing assistants
Best draft of a blog post or email in my voice?
6 candidates
- 13Not yet tested
Data analysis
Which AI analyses my spreadsheet correctly?
6 candidates
- 14Not yet tested
Workflow automation
What should I use to automate a workflow between apps with AI?
6 candidates
- 15Not yet tested
Local / open models
Easiest way to run a good model on my own machine?
6 candidates
The score, when we have one
FitScore = weighted average × 20, rounded to the nearest integer. The weighted average is the sum of (criterion score × weight) divided by 100.
- Output quality30
- Reliability15
- Cost and free tier15
- Speed10
- Setup friction10
- Agent, API, and MCP support10
- Privacy and data10
1 to 5, before the 0 to 100
- 5 Excellent. No fixes.
- 4 Good. One minor fix.
- 3 Usable. Several fixes.
- 2 Poor. Partial result.
- 1 Fails. Does not complete the task.
Weights sum to 100. We do not publish a FitScore from a single undocumented session.
Latest tested tools
- Claude Code94
Coding agents
- OpenCode93
Coding agents
- Codex CLI90
Coding agents
- Cline88
Coding agents
- Google Antigravity79
Coding agents
A weekly note for people who build with agents
New tests, new tools, and corrections. Free. Subscribing does not change a FitScore. Latest from the catalog: Coding agents: first provisional FitScores (5 Nov 2026).