Agent tools
Tools for running agents on an InferOptimal endpoint. Your coding agent lists them with query action list_agent_tools, asks how you want to deploy your agents and installs only what you choose.
Every tool
InferOptimal OpenEvolve tool Recommended
swarm tool
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.
Runs one large evolutionary search for you. Many agents propose versions of a program, an evaluator you write scores each one, and islands of candidates evolve toward the best.
Fits: A program, prompt or configuration with a score you can measure repeatedly and want to raise.
Guide:
node ~/.inferoptimal/agents.mjs guide openevolveMicrosoft Agent Framework fan-out Recommended
swarm tool
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.
Microsoft's open-source Agent Framework running our fan-out on your endpoint: 4 investigators by default, or any number you choose, each with its own line of inquiry and tools to search and read a folder, work in parallel, and an aggregator weighs their findings and writes the answer.
Fits: A hard question over a large folder of documents where several lines of inquiry should run at once, on an endpoint with room for heavy token use.
Install:
node ~/.inferoptimal/agents.mjs tool install agent-frameworkDeepSeek Harness on your endpoint Recommended
swarm tool
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.
DeepSeek's open-source agent harness, pointed at your endpoint: one agent works a task through in the harness's own agent loop with file and shell tools, and its workflow tool can fan the task out to many agents.
Fits: One hard task over a folder or repository that one persistent agent should work through, such as a research question over a set of documents.
Install:
node ~/.inferoptimal/agents.mjs tool install deepseek-harnessInferOptimal subagents
subagent tool
Your coding agent keeps its own model and hands whole coding tasks to agents running on the endpoint. They work in parallel in your project, and your agent reviews each result.
Fits: Many separate coding tasks at once: features, fixes, tests or research across a repository.
OpenResearch add-on for the OpenEvolve tool In development
swarm tool
In development: still offered, but not recommended yet, because it is not performing well enough. Its measured results are below.
Runs the OpenEvolve tool on a research question with no fixed metric, over a folder or repository. With --fanout, many single-track agents on the endpoint each do one directed job, such as reading a slice of the documents or hunting for a rule that overrides another, and report exact passages; a decider model rates what they found and writes the next brief.
Fits: A research question with no fixed metric over a large set of documents, especially one where the obvious answer may be overridden by something buried in the material, such as a later amendment, a definition or another document.
Install:
mkdir -p ~/.inferoptimal/openresearch && cd ~/.inferoptimal/openresearch && curl -fsSLO https://www.inferoptimal.com/openresearch/Makefile && make fetch install