fix(mcp-chat): loop tool calls until a final answer

[Ernstsen]

Enable multi-round tool calls for MCP-Chat plugin

Issue:

When an llm-request in the MCP-Chat required trailing tool-calls the response would look like this:

<|tool_call>call:catalog.query-catalog-entities{query:<|"|>Platform Engineering<|"|>}<tool_call|>

Cause:

processQuery ran a single round of tool calls, then made a follow-up LLM request WITHOUT tools. When a question needed more than one tool call, the model attempted another call in that tool-less follow-up; OpenAI-compatible servers don't parse tool calls when tools is absent, so the model's raw tool-call tokens leaked into the response content (and the follow-up content discarded the real answer).

Replace the single round with a capped agentic loop (max 8 iterations) that always passes the tools on every sendMessage call and executes returned tool calls until the model responds with a final text answer, with a graceful fallback if the cap is reached.

The per-call execute/ record logic is extracted into a private helper, for readability.

Adds tests covering the multi-round loop and the iteration cap.

✔️ Checklist

• A changeset describing the change and affected packages. (more info)
• Added or updated documentation
• Tests for new functionality and regression tests for bug fixes
• Screenshots attached (for UI changes)
• All your commits have a Signed-off-by line in the message. (more info)

[backstage-goalie]

Changed Packages

Package Name	Package Path	Changeset Bump	Current Version
@backstage-community/plugin-mcp-chat-backend	workspaces/mcp-chat/plugins/mcp-chat-backend	patch	v0.11.0

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

This PR fixes multi-round tool-calling behavior in processQuery by iterating through tool calls until the model returns a final answer (or a max-iteration cap is reached), preventing later rounds from losing access to tool definitions.

Changes:

• Reworked processQuery to loop across multiple LLM/tool-execution rounds with a max-iteration safeguard and a fallback reply.
• Extracted tool execution + conversation-history updates into a helper (executeAndRecordToolCall).
• Added tests validating tools are passed on every round and that max-iteration fallback triggers.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 3 comments.

File	Description
workspaces/mcp-chat/plugins/mcp-chat-backend/src/services/MCPClientServiceImpl.ts	Implements multi-round tool loop, iteration cap, and refactors tool execution/message recording.
workspaces/mcp-chat/plugins/mcp-chat-backend/src/services/MCPClientServiceImpl.test.ts	Adds regression tests for multi-round tool passing and max-iteration fallback behavior.
workspaces/mcp-chat/.changeset/fix-multi-round-tool-loop.md	Adds a patch changeset describing the multi-step tool-call fix.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.