Log inGet started

SQUAD / Multi-agent orchestration

Let your agents work as a Squad.

Give your lead agent a team. Delegate focused work, run independent tasks in parallel, and bring results back for review. Need to choose an approach first? Ask two agents to debate it.

Start your computer See how Squad works
  • Your agents & models
  • On your cloud computer
your-project / squadExample workflow
You set the direction

“Build this feature with a Squad.”

Lead agentPlan · delegate · collect
ResearchExplore the codebase
ImplementationBuild the feature
TestsCover the edge cases
Results return to the leadReview & iterate
Focused tasks. Independent work in parallel.
One lead to bring it together.

What is Squad?

Squad is TermSquad’s multi-agent orchestration capability for AI coding agents.

It connects agents around one project: a lead delegates focused tasks and collects results, or two peers debate a decision. Everything runs on your TermSquad Computer.

Two ways to work together

Delegate the work.
Debate the decision.

Choose the kind of collaboration your task needs. A focused team to execute. Two independent minds to explore an approach.

01 / DELEGATE

A clear plan.
More hands on the work.

A lead agent gives specialists a focused task, collects their results and coordinates review. Independent tasks can run at the same time.

EXAMPLE · BUILD A FEATURE

  1. Scope the taskLead defines the goal and file ownership
  2. Split independent workSpecialists research and implementParallel
  3. Review the completed changeA separate reviewer returns findings
  4. Bring the results togetherLead checks the outcome and next steps

Useful for builds, fixes, research and tests.

02 / DEBATE

A real trade-off.
Two independent perspectives.

Two equal agents form their own positions, then challenge each other’s reasoning. You get an agreed proposal or an explicit no-consensus result.

EXAMPLE · CHOOSE AN APPROACH

“Should this API use REST or GraphQL?”

Peer AIndependent position
Peer BIndependent position
Challenge assumptionsConsensus or no consensusTranscript and result saved on your computer

Starts when explicitly requested.

Parallel where it helps. Sequential where it matters. Tasks that depend on an earlier result wait for that work to finish.

How to use Squad

Your agents.
A shared direction.

Start with the tools you use, give each agent a clear role, and keep the lead focused on the outcome.

Explore supported agents

The TermSquad Skill connects the workflow.It gives supported agents instructions for delegation, review and working together on your computer.

  1. 01

    Choose your team.

    Install and connect supported agents on your TermSquad Computer. In Settings, choose the agent and model for implementation, review and each debate peer.

  2. 02

    Give the work a direction.

    Give your lead agent a clear task and ask it to use Squad. It can delegate independent work to specialists. Explicitly request a debate when you want two agents to examine a decision.

  3. 03

    Collect, review, continue.

    The lead collects output and status from delegated agents, checks the work and brings it through review. A debate returns an agreed proposal or a clear no-consensus result.

You set the boundaries

A capable team.
Still your call.

Choose who does the work, when delegation is welcome, and how your computer handles the load.

An agent and model for each role.

Choose your defaults for Implementation, Review, Peer A and Peer B in Squad Settings. Use supported agents you have installed and connected.

Delegation on your terms.

Let agents delegate proactively when useful, or turn Automatic delegation off and request it yourself. Debate always starts with an explicit request.

Room to work, with guardrails.

Parallel work depends on available resources. TermSquad Guard checks capacity and limits memory for heavy tests run through Guard. Squad reports blocked, stalled or timed-out work to the lead.

Explore workload protection

Your agents use your own subscriptions or API keys. AI usage is billed separately by your providers.

Frequently asked questions

Get to know
your Squad.

What runs in parallel, who chooses the agents, and how it all fits together.

Explore the documentation
What is Squad, and what is multi-agent orchestration?

Squad is TermSquad’s multi-agent orchestration capability for AI coding agents. It coordinates agents on your cloud computer: a lead delegates focused tasks and collects results, or two peers debate a decision.

When should I delegate, and when should I debate?

Delegate when the path is understood and the work can be split into focused tasks. Request a debate when the approach is uncertain and you want two equal agents to develop independent positions, challenge assumptions and examine the trade-offs.

Can agents work in parallel on the same project?

Yes, when their tasks are independent and the computer has capacity. Give agents separate file scopes, or let Squad give each implementer its own Git worktree for overlapping changes. Work that needs an earlier result, such as reviewing a completed implementation, runs in sequence.

How does Squad check the work?

A lead can declare the checks a task must pass, such as tests or a build. Squad runs them, with TermSquad Guard when it is enabled, and records the results so the lead can verify the evidence before accepting the work. Checks complement review; they do not prove a change is correct.

Can I choose different agents and models for each role?

Yes. Squad Settings let you choose an installed supported agent and model for Implementation, Review, Peer A and Peer B. Connect each agent with your own account or API key. The agent’s available models depend on its provider and setup.

Does Squad start other agents automatically?

The Automatic delegation setting lets supported agents use Squad proactively when useful. Turn it off to keep delegation explicit. You can still ask an agent to use Squad with this setting off. Debate always requires an explicit request.

What happens if agents disagree or a task gets stuck?

Debate reports consensus only when one peer explicitly accepts the other’s final proposal. Reaching the round limit produces a valid no-consensus result. Delegation returns status and output, including blocked, stalled or timeout information, so the lead can decide what to do next.

How is Squad different from Squad Memory?

Squad coordinates work between agents. Squad Memory stores durable project knowledge for agents and future sessions. Debate transcripts and results are saved locally; accepted decisions can be saved explicitly to memory, but debates do not write to Squad Memory automatically.

Is Squad a separate AI model or subscription?

Squad is a capability of your TermSquad Computer, not an AI model. You bring supported agents and their subscriptions or API keys. AI subscriptions and usage are not included in your TermSquad plan; running more agents can increase provider usage.

Explore all features
Squad MemoryShare durable project knowledge across agents and sessions.Persistent Agent SessionsDisconnect and return to the same agent workflows later.Built for Agent WorkloadsProtect the computer from bursty, concurrent agent workloads.
Go deeperOpen the docs hubSee the complete product workflow

One goal. A whole team.

Your next project.
With a Squad behind it.

Give your agents a computer, a role, and room to work together.

Start your computer Find your plan

AI subscriptions and usage are not included.