Major gaps and limitations with Intercom MCP | Community
Skip to main content
Submitted

Major gaps and limitations with Intercom MCP

  • March 16, 2026
  • 1 reply
  • 54 views
P

Hi, Sending some feedback on the MCP integration from a developer use case perspective. We've ran into some major gaps/roadblocks when using the MCP:
 

1. Expose created_at filtering in search_conversations. The underlying API supports date filtering but the MCP tool doesn't expose it. Agents scoped to a time window have to paginate all historical results and apply the cutoff manually.

2. Expose source.body text search in search_conversations. Again, supported by the API but not exposed in the MCP. Forces agents doing content-based retrieval to call get_conversation on every result individually.

3. Add source.subject filtering to search_conversations. Not sure if this exists in the underlying API — if it does, same ask as above. If not, it's a genuine gap: for email-based workflows, subject line is the most reliable signal for identifying conversation origin.

4. fetch drops source.subject from its output. get_conversation returns it correctly; fetch doesn't include it in its summary. Forces a redundant call for a consistently useful field.

5. Add tag filtering to search_conversations. Tags are the primary categorization mechanism but aren't queryable at search time.

6. Add custom attribute filtering to search_conversations. search_contacts supports this — parity in search_conversations would make agent-based segmentation significantly more practical.

 

    1 reply

    S

    +1 from the Stilla team. We've hit all of these same walls.

    Stilla is an AI teammate and agent builder platform which integrates with Intercom and tools like Linear, GitHub, Slack, Notion etc. We use the Intercom MCP to give our AI agents context from customer conversations. The gaps listed directly limit how useful those agents can be.

    Points 1 and 2 are especially painful for us. Without created_at filtering and source.body search, our agents end up doing expensive workarounds just to stay scoped to relevant conversations.

    We also integrate with tools like Pylon which expose richer filtering and querying over customer conversations — and the contrast makes the gaps in Intercom's MCP more apparent.

    Beyond the read gaps, we'd really love to see write capabilities too. The ability to reply to conversations, update contact attributes, add tags, or close tickets via MCP would unlock very important agentic workflows. Right now we're limited to agents that observe but can't act.

    Would love to see Intercom close these gaps. These aren't edge cases. They're quiet expected for any agent doing meaningful work with conversation data.

    Terms & ConditionsAccessibility statement