Quickstart: your first agent in 10 minutes.

From zero to a working AI agent that can read your repo, file PRs, and report status in your team's Slack — without writing a line of orchestration code.

01Overview

Nodeweaver agents are long-lived, scoped to a workspace, and able to call any tool you've connected — version control, ticketing, messaging, your custom MCP servers, or arbitrary HTTP. You give them a goal and the resources to pursue it; they break down the work, hand off subtasks to other agents, and return a verifiable trace of every model call, tool call, and decision they made along the way.

This quickstart walks through provisioning a workspace, adding the GitHub integration, and spawning a Repo Maintaineragent from the template library. By the end you'll have an agent that can triage issues, draft PR fixes, and post a daily summary to Slack.

02Prerequisites

• An active Nodeweaver beta invite (apply here if you don't have one yet)
• Admin access to at least one GitHub repository
• A Slack workspace where you can install apps
• Optional: the Nodeweaver CLI (npm i -g @nodeweaver/cli) for the terminal walkthrough

03Provision a workspace

Each Nodeweaver tenant can have multiple workspaces, each with its own File Cabinet, agent roster, and budget cap. Workspaces are the unit of access control — keep one per team or business function.

1. Sign in at app.nodeweaver.io and click New workspace in the sidebar.
2. Pick a name (e.g. platform-eng) and choose a region. Region affects model routing latency and data residency.
3. Set an initial monthly budget cap. The default is $500; you can change it any time.

Or do it from the CLI:

04Connect your first integration

Integrations are how agents reach the outside world. Each one is a typed API surface plus the credentials needed to call it. You connect once at the workspace level and any agent in that workspace can be granted access via its tool policy.

Install the GitHub app from the Integrations panel, pick the repos you want to expose, and accept the scopes. Agents will only see the repos you select — there's no path from the workspace to the rest of your org.

05Spawn an agent

The fastest path from zero is the template library. Templates bundle a system prompt, a default model, a starting tool policy, and a recommended budget. You can fork any template into your workspace and customize it.

To spawn the Repo Maintainer template programmatically:

// Spawn a Repo Maintainer agent in the platform-eng workspace
const agent = await nw.agents.create({
  workspace: 'ws_8f3a7b',
  template:  'repo-maintainer@1.2',
  name:      'platform-bot',
  model:     'claude-haiku-4-5',
  tools: [
    'github.repos.read',
    'github.pulls.write',
    'slack.dm.write',
  ],
  budget: { daily: 25, monthly: 500 },
  schedule: '0 9 * * 1-5',  // 9am weekdays
});

console.log(agent.id);  // → ag_5e2f1c

Or via the CLI:

06Open a thread

Threads are the durable conversation surface between you and an agent. They carry message history, intermediate tool calls, model traces, and any artifacts the agent produced. Open one from the dashboard, or fire off a request via API:

Example request

await nw.threads.open({
  agent:    'platform-bot',
  message:  'Triage the open issues filed since Monday and draft fixes for any that look like one-line config bugs.',
  priority: 'normal',
  stream:   true,
});

The agent will plan its work, post a checklist back into the thread, and start executing. You can watch every model call and tool invocation in real time, intervene with a follow-up message at any point, or hand off the thread to a teammate.

07What's next?

You've got a working agent. From here you might want to:

• Connect a second integration (Slack, Linear, Jira) to give the agent more context
• Add files to the workspace File Cabinet so the agent can read your design docs and runbooks
• Spawn a second agent and pair them with the Coordinator pattern
• Set up budget alerts and audit log exports for your security team