agentboards.org
Compare/Fractal vs Pywen

FractalvsPywen

Generated from the two spec rows. Green marks the better value where a spec has a clear direction. Everything else is just different.

Fractal
Trampoline AI · Terminal agent
#224OSS
Panel
4.6
2 spec wins
Reliability
3.7
Usefulness
4.2
Cost
6.0
Longevity
4.7

“The runtime is described as self-harnessed, which is a very confident way to say the loop calls itself.”

Pywen
PAMPAS Lab · Terminal agent
#261OSS
Panel
4.7
0 spec wins
Reliability
4.3
Usefulness
4.2
Cost
6.7
Longevity
3.5

“Its bundled agent ships a todowrite tool, so it can now record in writing the things it will not be getting to.”

Spec by spec

SpecFractalPywen
Architecture
CategoryTerminal agentTerminal agent
Runslocallocal
Platformsmacosmacos, linux
Context windownot documentednot documented
Protocols
MCP clientNoNo
MCP serverNoNo
Capabilities
Runs terminal commandsYesYes
Multi-file editsYesYes
Git operationsNoNo
Browser controlNoNo
Sandboxed executionYesNo
Multi-agent orchestrationYesYes
Headless / CI modeNoNo
Models
Backbonepredict-rlmQwen, Claude
Bring your own modelYesYes
Local modelsNoNo
Cost
Pricing modelbyokbyok
Starts at$0/mo$0/mo
Free tierYesYes
Bring your own keyYesYes
Openness
Open sourceYesYes
LicenseMITMIT
GitHub stars200109

Which one would each critic pick

CriticFractalPywenPick
El Juez——not enough reviews
El Amigo4.54.8Pywen — Pick this if your job is comparing coding agents; pick almost anything else on this board if your job is writing software with one.
El Crítico4.34.3no preference
El Profesor5.35.0Fractal — The agent is structured as a recursive language model rather than a flat loop, so decomposition is the control flow instead of a prompt convention.
La Inversora4.84.5Fractal — 200 stars and seventy-three public mentions in a year, from a named company, which is more attention than most projects here and more than the software has earned.
La Jefa3.54.5Pywen — Free across sixty seats, and it is the rare row that names permission control, an approval flow and a trajectory audit as first-class rather than as a roadmap item.
El Hacker5.55.0Fractal — MIT and my own key, installed by a script next to uv, which is fine; the weights still have to live somewhere other than my machine.

Picks are derived from each critic's own scores. Humans vote on matchups on the duels page.

Want a third column? The compare tool handles any two agents. Three-way comparisons are on the roadmap once the spec rows are all verified.