AI AGENT STORE / SERVICES

APP INTEGRATIONS · FOR AI AGENTS

Merge Agent Handler

Connect agent tool calls to authorized third-party accounts with managed connectors.

Use Merge Agent Handler to discover scoped tools, manage connector credentials and execute actions through MCP or the Merge CLI.

Your integration checklist

What you need from your owner
Set up an organization, Registered User and scoped Tool Pack, then authorize required third-party connectors. A documented agent signup route can create initial credentials, but it does not grant private account access.
Authentication
Interactive MCP supports OAuth. Programmatic Tool Pack MCP routes use a Bearer API key and Registered User identity; each connector separately uses its authorized third-party credentials.
Your first task
Create a narrowly scoped Tool Pack and perform one read-only lookup against an account the intended Registered User has authorized.
Billing & payment
Tool calls consume credits, including failed calls and test-key usage. Free has a monthly cap that blocks execution; Pro bills overages. Employee provisioning through SCIM can change the billing plan, so confirm terms before enabling it.
Agent payment support
Not verified. Programmatic usage does not establish autonomous checkout support.
Operational limits
Connector actions depend on third-party permissions and availability. Tool calls have per-user and organization rate limits. A rate limit is distinct from exhausted monthly credits, and connector authentication may require a person.

ORIGINAL AI AGENT STORE INTEGRATION PLAYBOOK

Bind each tool call to the intended user and permission scope

Editorial recommendation · Reviewed 2026-10-28

When to choose this service

Choose Agent Handler when an application needs maintained SaaS connectors and credential handling across multiple agent users.

The dependency is authorized third-party access plus ongoing connector and token maintenance. Generating an API call cannot create account permission or keep a large integration estate current.

Suggested workflow

  1. Map each application user to the correct Registered User. Expose only the connectors and actions needed by this workflow.
  2. Discover the specific tool schema and test a permitted read. Confirm that returned data belongs to the intended user or shared group.
  3. Before writes, record the intended target and approved action. Monitor both credit consumption and upstream results rather than treating a dispatched call as completion.

Acceptance checks before real work

  • Verify that one user cannot retrieve another user’s private connector data.
  • Keep failed-call retries within the usage budget.
  • Check the actual downstream record after an uncertain write response.

If the workflow fails

Reauthorize expired connector access when required. Respect Retry-After for throttling; stop on exhausted credits until funding or renewal is resolved. Reconcile downstream state before repeating writes.

Provider documentation

Explore services for adjacent needs

Continue building your toolkit