Hanzo

Agents

Build, deploy, and run autonomous agents — personas with a model, tools, guardrails, and shared memory.

Agents

Build, deploy, and run autonomous agents: configured personas with a model, a tool list, guardrails, and shared memory. Agents are wired into MCP for tools, Hanzo Brain for memory, and the LLM Gateway for inference. Tenant-scoped by org.

Quickstart

Define an agent and run it with the SDK. The same hk-* key authenticates every call.

import { Agent, Runtime } from '@hanzo/agents'

const agent = new Agent({
  name: 'researcher',
  model: 'claude-sonnet-4-5-20250929',
  brain: { org: 'startx' },              // shared memory
  tools: ['search', 'fetch', 'brain.recall'],
})

const result = await Runtime.run(agent, 'Summarize Q3 customer feedback.')
console.log(result.output)
from hanzo_agents import Agent, Runtime

agent = Agent(
    name="researcher",
    model="claude-sonnet-4-5-20250929",
    brain={"org": "startx"},
    tools=["search", "fetch", "brain.recall"],
)
print(Runtime.run(agent, "Summarize Q3 customer feedback.").output)

Run over the API

Agents also run over REST through the gateway. Point the SDK at https://api.hanzo.ai/v1 or call it directly:

curl -X POST https://api.hanzo.ai/v1/agents/runs \
  -H "Authorization: Bearer hk-..." \
  -H "Content-Type: application/json" \
  -d '{
    "agent": "researcher",
    "input": "Summarize Q3 customer feedback."
  }'

For the full REST shape — handoffs, guardrails, and orchestrators — see the API reference.

Observability

Every agent step emits OpenTelemetry spans and structured logs. Runs show up as traces so you can inspect each tool call, generation, and handoff, and score outputs through the evaluation surface.

  • Agents platform — orchestrators, guardrails, and self-hosting
  • MCP — the tool surface agents call
  • Traces — inspect and debug agent runs
  • Prompts — version the instructions agents use

How is this guide?

Last updated on

On this page