Multi-provider agent harness · MIT

One CLI that routes each task to the right model.

arccode picks a model by complexity, cost, performance, and intent, then spawns agents, loads skills, and drives tools, Anthropic, OpenAI, and local Ollama behind one interface.

pipx$pipx install arccode
pip$pip install arccode
script$curl -fsSL …/install.sh | sh
git$pipx install git+…/arccode
Quickstart → Source
What it does
Multi-provider
Anthropic, OpenAI, Ollama, and OpenRouter behind one normalized interface. Models are hot-swappable.
Smart routing
A classifier scores every model by capability, cost, and latency for the task. Explicit overrides always win.
Agents as files
Markdown with frontmatter, auto-discovered. Each agent pins a model policy, tool allowlist, and skills.
Skills on demand
Progressive disclosure: descriptions stay indexed, bodies load only when needed. Import or build new ones.
Orchestration
A coordinator spawns specialists and fans out subtasks, each routed to its own model.
Auth & more
OAuth login with token refresh, MCP servers, tool hooks, and resumable sessions.
The router
IntentModel tierWhy
bulk read / summarizesmallhigh volume, low stakes
implementworkhorsegood tools and code, moderate cost
design / debug / reviewfrontierneeds strong reasoning
chatsmalllatency matters

arccode whichmodel "…" explains any decision.

Quickstart
# install
pipx install arccode

# authenticate (API key, OAuth login, or run local with Ollama)
export ANTHROPIC_API_KEY=...
arccode auth login openai

# run a task, model auto-selected by the router
arccode run "Add a --json flag to the export command and test it"
arccode run "Design a rate limiter for 10k rps" --agent architect
arccode chat
arccode whichmodel "fix the failing build"