agentboards.org
Compare/Cowork Forge vs Manus

Cowork ForgevsManus

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

Cowork Forge
sopaco · Autonomous SWE
#258OSS
Panel
4.7
4 spec wins
Reliability
4.0
Usefulness
4.3
Cost
6.5
Longevity
4.0

“The engineer agent writes the code and then writes the delivery report, which is the most realistic simulation of a development team yet.”

Manus
Butterfly Effect (Manus) · Autonomous SWE
#205MCP
Panel
4.5
4 spec wins
Reliability
3.8
Usefulness
5.7
Cost
4.0
Longevity
4.3

“The free plan refreshes 300 credits a day, which is the agent equivalent of the hotel breakfast.”

Spec by spec

SpecCowork ForgeManus
Architecture
CategoryAutonomous SWEAutonomous SWE
Runslocalcloud, sandbox
Platformsmacos, linuxweb, macos, windows
Context windownot documentednot documented
Protocols
MCP clientNoYes
MCP serverNoNo
Capabilities
Runs terminal commandsYesYes
Multi-file editsYesYes
Git operationsYesYes
Browser controlNoYes
Sandboxed executionNoYes
Multi-agent orchestrationYesYes
Headless / CI modeNoYes
Models
BackboneClaude Code, Codex, OpenCodeManus 1.6, Manus 1.6 Max, Manus 1.6 Lite
Bring your own modelYesNo
Local modelsNoNo
Cost
Pricing modelbyoksubscription
Starts at$0/mo$20/mo
Free tierYesYes
Bring your own keyYesNo
Openness
Open sourceYesNo
LicenseMITproprietary
GitHub stars97n/a

Which one would each critic pick

CriticCowork ForgeManusPick
El Juez——not enough reviews
El Amigo5.05.5Manus — Pick Manus if the deliverable is a slide deck, a report and a small app from one prompt; pick Devin or Codex cloud if the deliverable is a pull request in a codebase that already exists.
El Crítico4.04.5Manus — Projects sync two-way to a private GitHub repo from a sandbox that has internet access, so an agent that browsed the wrong page has write access to your history.
El Profesor5.04.5Cowork Forge — Each role applies an actor-critic pattern for self-review, with human validation inserted at critical decision points. The pattern is named; the criteria the critic applies are not.
La Inversora4.54.8Manus — Meta's $2B plus acquisition announced in December 2025 was blocked by Chinese regulators in April 2026 and unwound, so the company you buy from is one that already sold itself once.
La Jefa4.04.3Manus — Team is $20 per seat, $1,200 a month for sixty before credits, concurrency is capped by tier up to 20 tasks at $200, and nothing on the page mentions SSO, audit logs or retention.
El Hacker5.83.3Cowork Forge — MIT and it drives Claude Code, Codex or OpenCode, so the subscription I already hold does the work. There is no MCP client, so my own servers never enter the picture.

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.