Intelligence that acts.
Macrodeep builds general agents that perceive, reason, plan, and act across open-ended environments — starting with the hardest one in business: go-to-market.
Language models describe the world. Agents change it.
The frontier is no longer a better predictor of the next token — it is a system that can be handed a goal and trusted to pursue it. Generality and agency are the same problem. An agent that can act across many environments has learned something a narrow tool never will.
We work end-to-end: the learning algorithms, the world models that let an agent imagine consequences before committing to them, the evaluation that tells us whether it is actually getting better, and the oversight that keeps it steerable as it does.
We chose go-to-market as our proving ground — research, outreach, qualification, and the judgment that connects them. It is open-ended, adversarial, and measured in real outcomes. An agent that can win here has learned to act.
Nico 2.5 is a large language model trained for go-to-market — research, outreach, qualification, and the judgment that ties them together. It doesn’t just draft. It acts.
01the flagship productNico
AI GTM engineerDescribe a motion in plain English — Nico plans it, builds the pipeline, and runs it across Salesforce, HubSpot, Clay, and the rest of your stack. Nothing sends without your approval.
Meet Nico →
02ai sales workforceSalestools.io
Powered by Nico 2.5An AI sales workforce. Autonomous SDR agents research, personalize, and engage prospects across email and LinkedIn — booking meetings on autopilot.
salestools.io ↗
03multi-agent command centergrain.sh
Powered by Nico 2.5A multi-agent command center. Hand it a goal in plain English; a team of named specialists researches, plans, executes, and ships a polished artifact.
grain.sh ↗411M verified contacts, 78M companies, real-time signals — exposed as an MCP server so AI agents query revenue data the way they query a tool.
many worlds.
We build agents that carry skills across tasks, tools, and domains — not a new model for every problem. Breadth is the point.
to action.
Reasoning only matters when it meets consequence. We train for long horizons, where an agent must plan, recover from error, and finish.
controllable.
Capability and oversight grow on one curve. We invest in evaluation and control so general agents stay legible as they get stronger.
We are not building a bigger model. We are building something that can act — and we intend to do it safely, in the open, and soon.
Come build the
first general agents.
Headquartered in San Francisco, with offices in New York and Bucharest. We’re hiring researchers, engineers, and people who ship. If long horizons keep you up at night, talk to us.