IQ
Open Source fogfish/iq brew install fogfish/tap/iq

IQ

AI workflow automation. Define multi-step agent pipelines in YAML — no code required.

IQ is a Go CLI and SDK for orchestrating LLM agent workflows. Write a YAML file that chains prompts, routes on LLM output, integrates external tools via MCP, and processes data at scale — then run it anywhere with a single command.

Capabilities

What it does

Declarative YAML Workflows

Define complex multi-step agent pipelines in simple YAML. Jobs contain steps that chain, route, and iterate — no coding required.

Multi-Provider LLM

Works with AWS Bedrock (Claude), OpenAI, and local models via LM Studio. Switch providers without touching your workflow definitions.

Intelligent Routing

CEL expressions for conditional branching. Classify content with an LLM, then dispatch to the right workflow branch automatically.

MCP Integration

Native Model Context Protocol support. Connect external tool servers for filesystem access, APIs, and specialized capabilities.

Batch Processing

Process directories and S3 buckets at scale. Resumable execution with retry strategies and error recovery built in.

Smart Caching

SHA256-based prompt caching with automatic invalidation. Reduce LLM costs and speed up iteration during development.

Architecture

How it works

IQ parses your YAML workflow, resolves job dependencies, and executes steps through a configurable LLM provider. Conditional routing uses CEL expressions evaluated against step outputs, allowing workflows to branch, loop, and converge based on model decisions.

IQ architecture diagram

Example

A workflow in YAML

Define jobs, chain steps, and route on LLM output — all in plain YAML. No framework to learn, no boilerplate to write.

# extract-and-summarize.yml
name: extract-and-summarize
jobs:
  main:
    steps:
      - name: extract
        uses: prompts/extract.md
        output: data

      - uses: prompts/classify.md
        switch:
          - when: choice == "technical"
            route: technical-summary
          - when: choice == "business"
            route: business-summary

  technical-summary:
    steps:
      - uses: prompts/tech-summary.md

  business-summary:
    steps:
      - uses: prompts/biz-summary.md

Usage

Run it in seconds

Install via Homebrew, point IQ at your workflow file and an input document. It handles the rest.

zsh — 80×24

$ brew install fogfish/tap/iq

 

$ iq agent -f extract-and-summarize.yml report.pdf

 

> Running extract...

> Classifying content... → technical

> Running technical-summary...

 

Done — output written to stdout

 

$ _

Installation

Get started

IQ is available via Homebrew and as open source on GitHub. MIT licensed.

Homebrew
brew install fogfish/tap/iq
GitHub
github.com/fogfish/iq

Source, issues, and full documentation

Try IQ

Start automating AI workflows

IQ is open source and ready to use. Star it on GitHub, read the docs, or reach out if you want to build something together.