agentboards.org
Compare/Cosine vs FuXi

CosinevsFuXi

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

Cosine
Cosine AI · Terminal agent
#100MCP
Panel
5.8
1 spec wins
Reliability
5.7
Usefulness
6.8
Cost
5.2
Longevity
5.5

“It used to be called Genie and is now named after a function that always comes back around, which is a bold choice for a startup.”

FuXi
FUXI · Terminal agent
#90MCP
Panel
5.8
5 spec wins
Reliability
5.0
Usefulness
6.5
Cost
7.3
Longevity
4.3

“More than fifty built-in tools in one binary, which is either thoroughness or the longest permission prompt in this category.”

Spec by spec

SpecCosineFuXi
Architecture
CategoryTerminal agentTerminal agent
Runslocal, cloud, sandboxlocal
Platformsmacos, linux, windows, webmacos, linux, windows
Context windownot documentednot documented
Protocols
MCP clientYesYes
MCP serverNoYes
Capabilities
Runs terminal commandsYesYes
Multi-file editsYesYes
Git operationsYesYes/commit and /review create commits, review code and open pull requests.
Browser controlYesBrowser tools drive Chrome or Chromium over the DevTools Protocol, which requires launching the browser with --remote-debugging-port and setting cdp_url in ~/.cosine/config.toml. Yes
Sandboxed executionYesRemote environments are isolated containers built from a Dockerfile; local runs are not sandboxed. No
Multi-agent orchestrationYesYes
Headless / CI modeNoYes`fuxi setup-token` is documented specifically for headless and CI use.
Models
BackboneLumen Scout, Lumen Outpost, GPT-5.5, Claude Sonnet 4.6, Claude Opus 4.7, Gemini 3.1 Pro, DeepSeek, Kimi K2.6, MiniMax M2.7, Qwenany OpenAI-compatible provider, Anthropic, Gemini, Amazon Bedrock, Google Vertex
Bring your own modelYesLimited to ChatGPT and Claude models accessed through your own provider account. Yes
Local modelsNoYesA custom base_url is documented through config, the FUXI_BASE_URL environment variable and the -b/--base-url flag, which covers a locally hosted OpenAI-compatible server.
Cost
Pricing modelsubscriptionbyok
Starts at$19/mo$0/mo
Free tierNoYes
Bring your own keyYesYou can sign in with your own ChatGPT or Claude Max subscription, which routes billing to that provider instead of consuming Cosine credits. Yes
Openness
Open sourceNoNo
Licenseproprietaryproprietary
GitHub starsn/a3,350

Which one would each critic pick

CriticCosineFuXiPick
El Juez——not enough reviews
El Amigo6.36.5FuXi — Pick it if you want one downloaded binary and no runtime to install; pick OpenCode when you would rather your terminal agent be something you can read.
El Crítico6.05.3Cosine — The same agent has two different safety models: remote environments are isolated containers, and local runs are not isolated at all.
El Profesor5.56.3FuXi — Classifying shell commands by parsing them into a syntax tree is the correct method; pattern matching on command strings is what everyone else does and it is defeated by quoting.
La Inversora6.05.3Cosine — Owning the Lumen models is the only durable asset here, and add-on credits sold at four to six dollars a million is where the margin is decided.
La Jefa5.35.0Cosine — Sixty engineers on the entry plan is $1,140 a month before a single credit is topped up, and there is no headless mode to put any of it in a pipeline.
El Hacker5.86.5FuXi — Closed source, but FUXI_BASE_URL or the -b flag points it at any OpenAI-compatible endpoint, which includes the server running on my own 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.