SuperAPI AI Directory
Partner中文AI newsFavorites
Skillssetup-agentkit

setup-agentkit

Agents & models

View repository

Updated 10-16 04:04

About this skill

Setup AgentKit

Give this project a dashboard connection, env credentials, and a first connector path. Then stop.

Guardrails

  • MUST wait for dashboard credential values. MUST NOT invent them.
  • MUST record the Connection Name exactly as the dashboard shows it (github-connect when the user named no connector).
  • MUST NOT write app-code tool calls. MUST name integrate-agentkit before waiting for credentials.

Gotchas

  • Read SDK credentials from SCALEKIT_ENVIRONMENT_URL, SCALEKIT_CLIENT_ID, and SCALEKIT_CLIENT_SECRET. Some samples use SCALEKIT_ENV_URL for the same URL; use SCALEKIT_ENVIRONMENT_URL here.
  • This skill stops after the connection. A connected account belongs to integrate-agentkit.
  • New environments ship one connection: GitHub, Connection Name github-connect. Every other connector, Gmail included, needs its own connection in the dashboard before the next skill.
  • Record the dashboard Connection Name. Later SDK calls use that exact string, not the connector slug.
  • The Scalekit MCP server at https://mcp.scalekit.com needs no extra env vars. It does not replace the SDK credentials.
  • Look up connectors at https://docs.scalekit.com/agentkit/connectors.md.
  • Wait for the user to supply credentials. Do not invent values.

Step 1 — Get API credentials

Open app.scalekit.com → Developers → Settings → API Credentials.

Collect:

SCALEKIT_ENVIRONMENT_URL=https://your-env.scalekit.com
SCALEKIT_CLIENT_ID=<from dashboard>
SCALEKIT_CLIENT_SECRET=<from dashboard>

Before you wait: record Connection Name github-connect when the user named no connector. Name integrate-agentkit as the next skill after they write .env. Then wait.

Done when: the three names are shown, Connection Name github-connect is recorded if no other connector was named, integrate-agentkit is named, and you are waiting for the three values.

Step 2 — Write project env

Put the three variables in the project's env file (.env or the existing env file). Keep secrets out of source.

Done when: the env file contains all three names, and source files do not hardcode the secret.

Step 3 — Create the first connection

Ask which connector to start with only when the user has not named one. Default is GitHub, which needs no setup: record Connection Name github-connect.

If the dashboard already has a row for the named connector, record that Connection Name exactly as shown.

Otherwise have the user create it:

Scalekit Dashboard → AgentKit → Connections → Add connection → select the connector → set Connection Name → Save.

For OAuth connectors that need a provider app, open the live guide: https://docs.scalekit.com/agentkit/connections/

Wait for the user to confirm the row. Do not invent the name.

Done when: a Connection Name that exists in the dashboard is written down (github-connect by default).

Step 4 — Record the first connector path

Keep these three items for the next skill:

Done when: those three items are known in this session.

Step 5 — Name the next skill and stop

If the user asked for OpenClaw, Hermes, or an always-on host, name integrate-agentkit-host.

Otherwise name integrate-agentkit. That skill creates a connected account, an authorization link, and one execute_tool call.

Do not write those app-code calls here.

Done when: integrate-agentkit or integrate-agentkit-host is named, and this skill has stopped.

Reach for

  • setup-scalekit if the plugin is missing
  • discover-connectors for the live tool catalog
  • integrate-agentkit-host for OpenClaw or Hermes
  • setup-saaskit if the user wanted app login

Live lookups