Developers

Quickstart

Install HiveBase for an AI client over MCP, or call the Platform API directly from a backend service.

1. Install the hosted MCP endpoint

Use this path for AI apps and coding agents. The client discovers OAuth 2.1 metadata, registers itself, and opens the browser consent flow.

claude mcp add hivebase \
  --transport http \
  https://mcp.hivebase.ai/

2. Verify the caller

After auth, ask the client to call auth.whoami. It returns the token kind, workspace, scopes, rate limit, and WU balance.

auth.whoami({})

3. Read context before work

Start with read tools. Reads are the adoption path: agents can inspect Brain, decisions, tasks, and account context before taking action.

brain.search({
  query: "Acme renewal status",
  limit: 5
})

4. Write back with confirmation

When an OAuth-authenticated agent proposes a ledger-affecting write, HiveBase returns confirmation metadata instead of silently committing the change.

decision.record({
  title: "Adopt short-lived OAuth tokens",
  judgment: "Use OAuth for MCP clients and hb_sk_* keys only for CI.",
  surface: "developer"
})