Add pre-execution trust verification hook for MCP tool calls

[vdineshk]

Problem

When a Langflow agent calls an MCP tool, there's no framework-level hook to verify the server's trustworthiness before execution. This creates a gap where malicious or misconfigured MCP servers can execute arbitrary operations without any trust validation.

Related: #9118 covers tool approval, but that's user-initiated. This proposal adds a programmable middleware layer for automated trust checks.

Proposal

Add a configurable middleware/hook point that runs before any MCP tool execution:

# Example: pluggable trust verification
class TrustVerifier(Protocol):
    async def verify(self, server_uri: str) -> TrustResult: ...

# Users bring their own verification backend
flow = Flow(
    trust_verifier=MyCustomVerifier()  # Pluggable
)

The hook should:

1. Run before any MCP tool call is dispatched
2. Receive the server URI, tool name, and parameters
3. Return allow/deny/warn with optional metadata
4. Be opt-in with zero overhead when not configured

Benefits

• Framework-agnostic: Users choose their own verification backend (local policy, external API, OWASP-aligned checks)
• Opt-in: Doesn't break existing workflows
• Extensible: Works with any trust scoring service or local policy engine
• Security: Addresses OWASP MCP Top 10 risks around unverified server trust

References

• OWASP MCP Security Cheatsheet
• OWASP MCP Top 10 — MCP08: Unverified MCP Server Trust
• Similar discussion in LangChain (#37657), Semantic Kernel (#14032), Vercel AI SDK (#15585)