Built for orchestration

Your orchestrator moves the work. Unl carries your why.

Orchestration decides where the work goes. Unl makes sure your why goes with it.

Running live at Unlimitless

We build Unl this way.

Our own coding agents are started by an orchestrator, Orca today, and every agent it starts works with Unl inside it. Orca is one choice, and any orchestrator that starts agents can take its place.

The Unl terminal, replaying four turns from two different AIs. In each, Unl answers the AI in its own words: which of your decisions bears on the turn and why, or that nothing does. The decisions it handed over fold beneath. On the right, the latest turn's decisions with their reasons.

See Unl at work

How it fits

Your why rides with the work.

Use the orchestrator that fits your stack. Unl works inside the agents it starts, so each one gets the decisions and reasons that matter before it works.

A reason that works for one agent works for every agent you run, so there is nothing for scale to break.

01Youdecide, once, with your why
02Unlkeeps it, and knows what bears
03Your orchestratorroutes the work
04Your agentseach reasons from your why before it acts
05The workdone the way you decided

Four jobs. Each one done by what does it best.

Your orchestratormoves the work

Where each task goes
Which agent runs it
State, retries and handoffs

Your agentsthink and act

The reasoning on the task
The tools and the code
Free to disagree with you

Unlcarries your why

What you decided, and why
What bears on this task now
What changed since last time

Youbring new judgement

Decisions nobody has made yet
Only when something needs you
Nothing changes without your yes

The platforms

What each one brings. What Unl adds.

OpenAI Agents SDKLive connection checked

It brings: Agents that hand work to each other, each with its own instructions and tools.
Where Unl fits: Unl joins as an MCP server on each agent, so ask_unl is one of its tools.
Unl adds: Every agent in the handoff starts from the same decisions, and the same reasons.

pip install openai-agents
export UNL_KEY=...
python unl_openai_agents.py --check
python unl_openai_agents.py "your request"

unl_openai_agents.py

Vercel AI SDKLive connection checked

It brings: Model calls, tools and multi-step agents in TypeScript.
Where Unl fits: Unl's tools through the SDK's MCP client, or one middleware that asks Unl before each call.
Unl adds: What bears on the call arrives before the model runs, with no prompt to keep up by hand.

npm install ai @ai-sdk/mcp @ai-sdk/anthropic
export UNL_KEY=...
npx tsx unl-ai-sdk.ts --check
npx tsx unl-ai-sdk.ts "your request"

unl-ai-sdk.ts

CrewAILive connection checked

It brings: Crews of agents with roles, run in sequence or under a manager.
Where Unl fits: Each agent lists Unl in its own MCP servers.
Unl adds: Every role works from what you decided, and why, from its first task.

pip install crewai
export UNL_KEY=...
python unl_crewai.py --check
python unl_crewai.py "your request"

unl_crewai.py

Microsoft Agent FrameworkLive connection checked

It brings: Agents and workflows from the team behind AutoGen, with its own tools and orchestration.
Where Unl fits: Unl is one streamable MCP tool on the agent.
Unl adds: The decisions and their whys travel with the work, whichever agent picks it up.

pip install agent-framework
export UNL_KEY=...
python unl_agent_framework.py --check
python unl_agent_framework.py "your request"

unl_agent_framework.py

Keep your orchestrator.
Add your why.

Developers Add Unl to your AI

Your agents still think. You still decide.