Why

FastMCP was pinned to 3.3.1 while each minor upgrade was validated against the MCP transport matrix. The current 3.4.4 release passes the repository's MCP smoke and transport-routing coverage, so the exact pin can be relaxed while retaining a v3 compatibility boundary.

What Changed

• replace the exact fastmcp==3.3.1 requirement with fastmcp>=3.3.1,<4
• regenerate the root lockfile with uv lock --upgrade
• update FastMCP and FastMCP Slim from 3.3.1 to 3.4.4
• update the newly available transitive jaraco-functools release from 4.5.0 to 4.6.0

Implementation Details

The <4 upper bound remains intentional: this allows future compatible FastMCP v3 releases without silently adopting a breaking major version. The root dependency set had already received a broad refresh on main, so the regenerated lockfile only changed FastMCP, FastMCP Slim, and jaraco-functools, plus FastMCP's newly declared extras.

Testing

Passed:

• uv lock --check
• just test-smoke — 1 passed
• uv run pytest test-int/cli/test_routing_integration.py -q --no-cov — 32 passed, including stdio, streamable HTTP, and SSE routing
• just package-check — Claude, Codex, shared skills, Hermes, and OpenClaw package checks passed
• git diff --check

Broad suite result:

• just fast-check ran the full suite after dependency changes invalidated testmon: 3,399 passed and 43 skipped
• its only failure was the live postgres-openai semantic benchmark receiving 429 insufficient_quota from the configured OpenAI account; all FastMCP and local deterministic tests passed

Additional doctor result:

• just doctor completed initialization, API write, file sync, vector indexing, search, and clean-status checks
• cleanup then hit the existing sole-default-project guard: Cannot delete default project ... This is the only project in your configuration.

Risks / Follow-ups

The primary risk is behavioral drift in future FastMCP v3 minors. The existing MCP smoke, tool contract, integration, and transport-routing coverage remain the guardrail. FastMCP v4 stays excluded pending an intentional compatibility review.