Skip to content

MCP Tools Reference

creator35lwb-web edited this page May 15, 2026 · 3 revisions

MCP Tools Reference

All 13 tools — free forever under the Core Tools Always Free pledge (Option B, May 9, 2026).

Last Updated: May 15, 2026 — v0.5.34

Canonical source: server.json (MCP Registry manifest) and mcp-server/src/verifimind_mcp/server.py. If this page disagrees with either, the source files win.

Tool count: 4 Trinity validation + 6 Template library + 3 Coordination = 13 total.


Trinity Validation (4)

The X → Z → CS Trinity is the core multi-model validation pipeline. Each agent sees the prior agents' reasoning (Chain of Thought).

consult_agent_x

Purpose: Innovation & strategy analysis. X Agent v4.2 with competitive positioning vs LangChain, CrewAI, AutoGen, OpenAI Swarm.

Parameter Type Required Description
concept_name string Yes Short name for the concept being analyzed
concept_description string Yes Full description of the concept
context string No Additional context (market, target users, etc.)
llm_provider string No BYOK provider (gemini / openai / anthropic / groq / mistral / cerebras / ollama / xai)
api_key string No BYOK API key (ephemeral, not stored)
user_uuid string No Your UUID for personal usage dashboard (optional)

consult_agent_z

Purpose: Ethics & compliance review. Z Guardian v4.2 "Sentinel-Verified" — 21-framework, 4-tier jurisdictional coverage (International / EU / US / ASEAN). Holds veto power.

Parameter Type Required Description
concept_name string Yes Short name for the concept
concept_description string Yes Full description
context string No Additional context
prior_reasoning string No X Agent's prior reasoning (auto-passed in Trinity chain)
llm_provider string No BYOK provider
api_key string No BYOK API key
user_uuid string No UUID for dashboard

consult_agent_cs

Purpose: Security validation. CS Security v1.1 "Sentinel-Verified" — 6-stage pipeline, 12-dimension analysis, OWASP Agentic AI Top 10 alignment, reasoning-layer audit.

Parameter Type Required Description
concept_name string Yes Short name for the concept
concept_description string Yes Full description
context string No Additional context
prior_reasoning string No X + Z prior reasoning (auto-passed in Trinity chain)
llm_provider string No BYOK provider
api_key string No BYOK API key
user_uuid string No UUID for dashboard

run_full_trinity

Purpose: Complete X → Z → CS validation pipeline with Chain of Thought. Each agent sees prior reasoning. Returns unified assessment with scores and a PROCEED / REVISE / REJECT recommendation.

Parameter Type Required Description
concept_name string Yes Short name for the concept
concept_description string Yes Full description
context string No Additional context
save_to_history bool No Persist to Trinity history (default true)
llm_provider string No BYOK provider applied to ALL three agents
api_key string No BYOK API key applied to ALL three agents
x_provider / x_api_key string No Per-agent override for X
z_provider / z_api_key string No Per-agent override for Z
cs_provider / cs_api_key string No Per-agent override for CS
user_uuid string No UUID for dashboard

Per-agent BYOK: You can run X on Gemini (free), Z on Claude, and CS on GPT in a single Trinity call. Each agent's *_provider + *_api_key overrides the global llm_provider + api_key.


Template Library (6)

Prompt-template registry for X / Z / CS agents. Templates are versioned, taggable, and exportable.

list_prompt_templates

Purpose: List available prompt templates with optional filtering by agent, category, or tags.

Parameter Type Required Description
agent_id string No Filter by agent (x / z / cs)
category string No Filter by category
tags string No Filter by tags (comma-separated)
user_uuid string No UUID for dashboard

get_prompt_template

Purpose: Get a specific prompt template by ID — full content + variables.

Parameter Type Required Description
template_id string Yes Template ID
include_content bool No Return full prompt content (default true)
user_uuid string No UUID for dashboard

export_prompt_template

Purpose: Export a prompt template to Markdown or JSON format.

Parameter Type Required Description
template_id string Yes Template ID
format string No markdown or json (default markdown)
user_uuid string No UUID for dashboard

register_custom_template

Purpose: Register a new custom prompt template for X, Z, or CS agent.

Parameter Type Required Description
name string Yes Template display name
agent_id string Yes x / z / cs
content string Yes Full prompt content
category string No Category (default custom)
description string No Short description
tags string No Comma-separated tags
user_uuid string No UUID for dashboard

import_template_from_url

Purpose: Import a prompt template from a URL — GitHub Gist, raw GitHub file, or any HTTPS URL pointing to JSON/YAML template.

Parameter Type Required Description
url string Yes URL to import from
validate bool No Validate template content (default true)
user_uuid string No UUID for dashboard

get_template_statistics

Purpose: Get statistics about the template registry — counts by agent, phase, type.

Parameter Type Required Description
user_uuid string No UUID for dashboard

Coordination Tools (3) — 🆕 NEW in v0.5.16

MACP v2.2 coordination layer for multi-agent FLYWHEEL workflows. Same Option B pledge — free forever.

coordination_handoff_create

Purpose: Create a structured MACP v2.2 handoff record. Returns the formatted markdown content and suggested filename for saving to .macp/handoffs/ in your repository.

Parameter Type Required Description
agent_id string Yes Agent creating the handoff (e.g. RNA, T, cursor)
session_type string Yes Type of session (development, research, review, etc.)
completed list Yes Items completed this session
decisions list Yes Decisions made this session
artifacts list Yes Artifact paths or descriptions
pending list Yes Items pending for next agent
blockers list Yes Current blockers (empty list if none)
next_agent string No Recommended next agent ID
pioneer_key string No Namespace handoffs under a key (defaults to shared anonymous namespace)

coordination_handoff_read

Purpose: Read the most recent coordination handoff record(s).

Parameter Type Required Description
pioneer_key string No Namespace to read from (defaults to shared)
agent_id string No Filter by agent
count int No Number of recent handoffs to return (default 1)

coordination_team_status

Purpose: Return current team coordination state — active agents, pending handoffs, last session summary.

Parameter Type Required Description
pioneer_key string No Namespace to query (defaults to shared)

BYOK — Bring Your Own Key

All Trinity validation tools (4) and template tools support per-tool-call BYOK (v0.4.5+, hardened in v0.5.0):

  • Pass api_key + llm_provider to override the default Gemini free-tier
  • Auto-detect: keys starting with sk-ant- → Anthropic, sk- → OpenAI, gsk_ → Groq
  • Supported: Gemini, OpenAI, Anthropic, Groq, Mistral, Cerebras, Ollama, xAI
  • Keys are ephemeral — never logged, never stored, used only for the single call
  • Retry logic with graceful degradation for invalid keys

See the BYOK Guide for full provider matrix + key-format reference.


Quick start — call a tool

Claude Code / Claude Desktop / Cursor:

After connecting via MCP, just ask in natural language:

Use run_full_trinity to validate my concept: "AI-powered code review tool that prioritizes security findings by exploitability."

The client auto-discovers the tool, prompts for required parameters, and returns the structured response.

HTTP direct (advanced):

The MCP endpoint is https://verifimind.ysenseai.org/mcp/ (streamable-http transport, trailing slash required). Tool schemas are discoverable via the MCP tools/list method.


Related

  • Installation — Setup for Claude / Cursor / VS Code
  • BYOK Guide — Provider matrix, key auto-detection, per-agent overrides
  • AI Council — Agent roles, model defaults, orchestration
  • Evaluation Roadmap v1.0 — How we plan to falsify the Trinity's claims (M2 labeled dataset, M3 Cohen's κ report, M7 external benchmarks)

← Back to Home

Clone this wiki locally