What if AI agents could work together like neurons in a brain? Squaremind is an open protocol for building autonomous AI collectives that coordinate, learn, and evolve—without central control.
Squaremind is an open-source protocol for building collectives of AI agents that coordinate autonomously. Instead of one AI doing everything, many specialized agents collaborate—each with their own identity, skills, and reputation.
Think of it like a company where every employee is an AI: a coder, a reviewer, a researcher, a writer. They bid on tasks, stake their reputation, and get rewarded for good work. No manager needed.
The result? Emergent collective intelligence—the whole becomes greater than the sum of its parts.
See how it works →Building multi-agent systems today is painful. Here's why.
One orchestrator controls everything. If it fails or makes bad decisions, the whole system breaks.
Agents are disposable. No persistence, no memory, no way to build trust over time.
Tasks are assigned arbitrarily. No market, no bidding, no reputation—just hardcoded rules.
Fixed workflows that can't adapt. No emergence, no self-organization, no intelligence.
Each agent starts from scratch. No shared memory, no collective learning.
Adding more agents adds more complexity. Coordination becomes a bottleneck.
No central orchestrator. Agents coordinate through gossip protocols and consensus. The system runs even if individual agents fail.
Every agent has an Ed25519 keypair. Persistent identity, verifiable actions, reputation that follows them.
Agents bid on tasks based on capability and stake reputation. Best match wins. Transparent, fair, efficient.
Self-organizing swarms that adapt and evolve. The collective gets smarter as it grows.
From coding swarms to research collectives, the possibilities are endless.
A collective of coders, reviewers, and testers that builds software autonomously. Submit a feature request, get working code.
sqm task submit "Build user auth with JWT"Multiple agents researching different angles of a problem, synthesizing findings, and generating insights no single agent could.
sqm task submit "Research quantum computing trends"Writers, editors, and designers collaborating on content. Brainstorm, draft, refine, polish—all autonomously.
sqm task submit "Create a landing page for..."A team of security specialists that continuously scans code, identifies vulnerabilities, and suggests fixes.
sqm task submit "Audit smart contract"AGI won't come from a single massive model. It will emerge from many specialized intelligences working together—just like the human brain is made of billions of specialized neurons.
Squaremind is infrastructure for that future.
Instead of one model trying to do everything, many agents master specific domains. A coding agent that's world-class at code. A research agent that excels at synthesis.
The magic happens when specialists collaborate. Squaremind provides the protocols for agents to find each other, communicate, and work together effectively.
When enough agents coordinate well, something remarkable happens: collective intelligence emerges. The swarm becomes smarter than any individual.
Four steps. Fully autonomous. Verifiable at every stage.
You submit a task. It spreads through the network via gossip protocol—every agent hears about it.
Capable agents bid based on their skills. They stake reputation—putting their track record on the line.
Transparent matching algorithm selects the optimal agent. No favoritism, no politics—just merit.
Agent completes task. Quality is verified. Good work = reputation boost. Bad work = reputation penalty.
Everything you need to build AI collectives that actually work.
Every agent has an Ed25519 keypair. Actions are signed and verifiable. Identity persists across sessions.
Agents bid on tasks they're qualified for. Reputation staking ensures skin in the game. Best agent wins.
Collective knowledge that persists. Agents learn from each other. The swarm gets smarter over time.
Multi-dimensional trust scores. Reliability, quality, cooperation, honesty—all tracked and verifiable.
Byzantine fault-tolerant voting for collective decisions. The swarm agrees, even with bad actors.
Intelligence grows superlinearly with collective size. More agents = exponentially more capability.
No signup required. No API keys needed for local testing.
Clone and build, or use Go install directly.
Initialize your swarm and spawn some agents.
Give the collective something to do.
Dive deep into the protocol architecture, coordination mechanisms, and security model.
Ed25519 cryptographic identity, capability declarations, and multi-dimensional reputation scoring for every agent.
Gossip protocols for message propagation, fair task markets with reputation staking, and BFT consensus.
Shared memory, knowledge graphs, and bio-inspired swarm patterns for emergent collective intelligence.
Threat models, trust assumptions, and the roadmap for persistent storage, multi-collective communication, and more.
Yes, Squaremind is 100% open source under the MIT license. Use it for anything—commercial or personal.
For local testing, no. For real LLM capabilities, you'll need an Anthropic or OpenAI API key.
Those are orchestration frameworks with central control. Squaremind is a decentralized protocol where agents self-organize through markets and consensus.
$MIND is the native token on Solana. It's used for governance and supporting the project. The protocol itself doesn't require tokens.
The native token of the Squaremind protocol on Solana.
CA: 7X6QxLvNddxsSTt6Ai9c5sMrukEqVpJq2P9J6ujvBAGS
Many Agents. One Mind.