Built for your endpoint
Agent tools.
Choose how your agents work. Use the dedicated capacity you already have.
Compare measured harness results →Start here
One setup.
Your choice of tools.
Download the agents script once, then choose a tool below. Install only what you need; ordinary inference needs none of them.
Run in your terminal · Node 20 or later
mkdir -p ~/.inferoptimal && curl -fsSL https://infraoptimal.com/agents/inferoptimal-agents.mjs -o ~/.inferoptimal/agents.mjsAlready connected through MCP?
Ask your agent to call query action get_codex_config. It returns the same download and install in one paste.
Choose your workflow
Find the tool for the task.
Recommendations and limitations come from our retained harness results.
- swarm tool Recommended
InfraOptimal OpenEvolve tool
A program, prompt or configuration with a score you can measure and want to raise.
Read the guide ↗Install commandnode ~/.inferoptimal/agents.mjs tool install openevolveRecommendation & limitations
Recommended for well-defined tasks with a reward that can be hill-climbed: a score you can compute for every candidate and want to raise.
View measured results → - swarm tool Recommended
Microsoft Agent Framework fan-out
A hard question over many documents, worked as parallel lines of inquiry; uses many tokens.
Read the guide ↗Install commandnode ~/.inferoptimal/agents.mjs tool install agent-frameworkRecommendation & limitations
Recommended as the best overall harness we measured, with known failure modes where it fails completely: document retrieval, finding vulnerable files, a run that ends at its time cap with no answer, and very heavy token use.
View measured results → - swarm tool Recommended
DeepSeek Harness on your endpoint
One hard task over a folder or repository for one persistent agent.
Read the guide ↗Install commandnode ~/.inferoptimal/agents.mjs tool install deepseek-harnessRecommendation & limitations
Recommended as the safer alternative harness. That is our recommendation, resting on measured facts: one agent ended each of its short-task runs in minutes, none at the time cap, and was right 5 of 9 times on hard research. Its own measured weakness: one agent can run into the output limit on a long answer and end with none.
View measured results → - subagent tool
InfraOptimal subagents
Many separate coding tasks at once: features, fixes, tests or research across a repository.
Open the guidenode ~/.inferoptimal/agents.mjs guide subagents - swarm tool In development
OpenResearch add-on for the OpenEvolve tool
A research question with no fixed metric; a buried detail may override the obvious answer.
Read the guide ↗Install commandnode ~/.inferoptimal/agents.mjs tool install openresearchRecommendation & limitations
In development: still offered, but not recommended yet, because it is not performing well enough. Its measured results are below.
View measured results → - swarm tool
InfraOptimal web tools
A swarm that collects data from the web: searching, reading and extracting across many sources.
Install commandnode ~/.inferoptimal/agents.mjs tool install web --tier <read|collect|browse>