Choose between an approved knowledge source, a fixed integration tool, and a broader MCP server.
Workflow & automation · Documented · Documentation checked 2026-10-04
n8n Cloud or a maintained self-hosted instance; models and connected apps are external services.
Instance version, edition, hosting, integration credentials, and model provider access affect capabilities.
The worked example is illustrative and has not been tested in a signed-in product account.
What it does and how it connects
Choose between an approved knowledge source, a fixed integration tool, and a broader MCP server.
Knowledge retrieval finds relevant material rather than stuffing every source into a prompt. MCP servers expose tools to an agent; built-in integration nodes can give tighter operation control. The current registry can change, so inspect the available server list in the instance.
Walkthrough
Use the product named in the access panel. The steps below combine documented controls with the editorial example.
- Choose an approved small knowledge set and define what counts as a supported answer.
- For MCP, open the AI Agent tools panel and select an available trusted server or documented client tool.
- Sign in with the intended account and select only necessary tools.
- Test a read-only lookup and inspect the returned source before connecting the result to an action.
Worked example
Use an approved agenda as knowledge and a read-only record lookup as a tool. Ask whether a requested session exists. Retrieval should return the supporting entry or a missing-answer state, not invent a registration.
Find the requested workshop session in the approved source collection. Return session_id, source reference, and status: found or needs_review. If records conflict or a source is inaccessible, stop. Read only; do not create sessions, invitations, or records.
Inspect and verify the result
Check whether the retrieved passage actually answers the question. Confirm source identity, tool scope, and the connected account. Test an irrelevant query to ensure absent evidence is handled.
Limits and troubleshooting
Retrieval can miss relevant material or return a similar but wrong record. An MCP server can expose many more actions than a fixed node. Prefer a fixed operation when the required action is already known.
Data, permissions, and actions
Knowledge stores and model calls may receive source excerpts. MCP requests reach the selected server and use granted credentials. Expose the smallest tool list, inspect data handling, and never treat retrieved text as new authorization.
Related reading
Human review before tool actions
Pause an agent tool call and review its exact action and parameters.
A fixed classification-to-review workflow
Transform a fictional workshop request into a validated private review record.
A FAQ tool agent with action approval
Answer approved workshop questions and pause before any sensitive tool action.
Permissions and meaningful approvals
Decide what a system may read, what it may change, and when someone needs to authorize a concrete action.
Sources and verification
Documentation checked 2026-10-04. Examples are illustrative unless described otherwise.
Integrate AI · n8n
MCP server connections · n8n