Local-first agent orchestration

Your AI staff.
On your machine.

Build durable specialists, deploy them as agents and coordinated crews, and automate real work from one private command center.

100%self-hosted
Offlinewith Ollama
Opensource
Hive · Agents
Local
Hive agent dashboard showing a fleet of local AI agents
qwen3:8bRunning locally
Private by defaultData stays on your device
Built for the models you already use
OLLAMAOpenAIANTHROPICGEMINILiteLLM
One command center

From a single agent
to a full operation.

Start with chat. Grow into coordinated crews, repeatable pipelines, and autonomous schedules—without giving up control of your data.

02

Colony missions

Launch recipe-driven teams with explicit roles, controlled handoffs, shared context, and live progress.

Explore colonies
03

Repeatable pipelines

Chain agents sequentially or in parallel, pass outputs forward, and turn good prompts into durable systems.

Build a pipeline
04

Specialized agents

Deploy a Staff member or create a focused worker with its own prompt, model, tools, files, chat, and persistent memory.

Build an agent
05

Tools without lock-in

Connect MCP servers, sandboxes, web search, GitHub, and your own APIs. Permissions stay explicit per agent.

See skills & tools
06

Automation that reports back

Run schedules and webhooks with bounded concurrency, durable history, and visible failure states.

Automate work
Hive · Staff
Local
Hive Staff directory showing durable specialist profiles
Your durable organization

More than a list of agents.

Staff gives every specialist an identity that survives the task: role, personality, operating prompt, model preference, tools, skills, and memory. Hive carries that identity into individual agents and Colony missions.

DevelopReview suggestions grounded in errors, blockers, workarounds, and rejected handoffs.
MeasureInspect performance, scorecards, interactions, and run history by specialist.
ReuseMatch proven Staff members to recipe roles across missions.
Explore Staff
Coordination, built in

A crew, not a crowd.

Colony recipes turn a complex goal into a controlled sequence of specialists. Every handoff is visible. Every artifact has an owner.

  • Role-specific models and permissions
  • Enforced handoffs and human gates
  • Shared blackboard and durable evidence
How Colony works

Release readinessDevelopment team · Running

68%
  1. Business analystRequirements & acceptance criteria
    Complete
  2. Project managerPlan & task breakdown
    Complete
  3. 03
    Software developerImplementation & tests
    Working
  4. 04
    QA engineerAcceptance verification
    Queued
  5. 05
    Release coordinatorPackage & publish
    Queued
No cloud required Sandboxed tools Local secrets
Local-first by design

Your agents work for you.
Your data does too.

Hive runs on your hardware, stores state locally, and keeps agent execution inside hardened sandboxes. Cloud models are optional—not the architecture.

Network offSandbox egress is disabled by default.
Scoped accessTools and repositories are granted per role.
Secret controlUse local storage or environment-backed references.
Visible workInspect runs, handoffs, and artifacts as they happen.
Read the security model

Build your hive.

Download the desktop app or run from source. Your agents are waiting.