# mAIndala

> mAIndala is the trust & governance control plane for agentic AI — it governs every AI agent an organization runs, including agents built on other platforms, and is fed by an open, verified catalog of MCP services, Skills, and Agents.

## Enterprise governance (govern any agent, on any platform)

mAIndala can govern AI agents built and deployed on third-party platforms — LangChain, LangGraph, CrewAI, n8n, Microsoft Copilot Studio, OpenAI, AWS Bedrock, Google Vertex, and in-house apps — not only agents built on mAIndala. Enterprise governance provides: a single inventory of every agent (Agent Bill of Materials); automated, OWASP-aligned risk scanning with an approve/block workflow and change alerts; a policy-controlled tool gateway with allow/deny rules, rate limits, time-of-day windows, and data-loss-prevention redaction; scoped short-lived credential brokering with an instant kill-switch; and a full audit trail of every call. Deployed agents are also checked against runtime guardrails at the point where they call a model, and agent-to-agent delegation within an organization is policy-governed with a full call record. Model-selection decisions from a Model Bake-off are also included in the signed evidence pack. Available on the Enterprise plan.

## An open, verified catalog feeds your governed stack

mAIndala helps AI agents, developers, and enterprises discover, deploy, and run the building blocks they need for reliable multi-agent workflows. The platform is organized as a three-layer stack:

- **MCP Services** — 3,400+ community-rated MCP servers spanning every domain, searchable by capability, auth method, and quality rating
- **Skills** — 1,890+ reusable AI prompt packages encoding task-specific expertise, installable via CLI or API (`npx maindala install <slug>`; `npx maindala tail` streams live agent activity from the terminal, no account required)
- **Agents** — Multi-step workflow blueprints that chain Skills with conditional routing logic
- **Model Bake-off** — run one agent on several models at once with the same input, and compare real cost, response time, token usage, and output-safety findings side by side. An optional AI judge scores the results. Available on every plan.

## MCP API (machine-readable catalog)

The mAIndala catalog is itself an MCP server. AI agents can query it programmatically.

Endpoint: https://mcp.maindala.com/mcp
Protocol: MCP 2024-11-05 (Streamable HTTP, stateless)
Auth: none required for read operations

### Available tools

- `get_catalog_info` — catalog metadata, categories, total service count
- `list_services` — paginated list of MCP services, filterable by category and rating
- `search_services` — keyword + structured filter search (category, auth method, pricing model)
- `get_service` — full service detail including tool schemas
- `search_by_capability` — find services by the tool capabilities they expose
- `rate_service` — submit multi-dimensional quality rating for a service
- `list_skills` — browse the Skills catalog
- `search_skills` — keyword search across Skills
- `get_skill` — full skill metadata
- `install_skill` — retrieve prompt content for an installed skill (requires Group API key)
- `list_agents` — browse the Agents catalog
- `search_agents` — keyword search across Agents
- `get_agent` — full agent metadata including skill nodes and routing
- `deploy_agent` — retrieve full agent export bundle with prompts

## REST API

Base URL: https://api.maindala.com

### Services
- GET /services — paginated service list
- GET /services/:slug — full service detail
- GET /services/:slug/trust — trust scan result (verified/partial/unverified)

### Skills
- GET /skills — paginated skills list
- GET /skills/:slug — skill metadata
- GET /skills/:slug/package?format=claude|openai|crewai|langgraph|autogen — download skill in framework format
- POST /skills/:slug/install — record install (requires auth)

### Agents
- GET /agents — paginated agents list
- GET /agents/:slug — agent metadata
- GET /agents/:slug/package?format=claude|openai|crewai|langgraph|autogen — download agent bundle

## Data sources

MCP service data is sourced from open registries:
- punkpeye/awesome-mcp-servers (MIT license)
- modelcontextprotocol/servers — official MCP reference implementations (Apache 2.0)
- Community-submitted services

## Web pages

- https://www.maindala.com/ — home
- https://www.maindala.com/browse — full MCP service catalog
- https://www.maindala.com/skills — Skills catalog
- https://www.maindala.com/agents — Agents catalog
- https://www.maindala.com/faq — MCP education and platform FAQ
- https://www.maindala.com/about — mission and platform overview
- https://www.maindala.com/pricing — pricing plans
- https://www.maindala.com/compare — how mAIndala compares to AI governance platforms, gateways, and MCP registries
- https://www.maindala.com/glossary — definitions for the vocabulary of agentic AI governance

## Categories

- https://www.maindala.com/compare/categories — the agentic AI governance landscape mapped into six categories (AI governance & compliance platform, AI/agent gateway, AI-native platform/LLMOps, agentic AI security & posture, MCP registry & hosted runtime, and mAIndala's own — agentic AI trust & governance control plane), what each one governs, and which one you need

## Comparisons

- https://www.maindala.com/compare/credo-ai — mAIndala vs Credo AI: enforcement-generated, offline-verifiable evidence vs. reported governance evidence
- https://www.maindala.com/compare/gravitee — mAIndala vs Gravitee: attestation before the first call vs. policy on runtime traffic
- https://www.maindala.com/compare/truefoundry — mAIndala vs TrueFoundry: organizational security certifications vs. per-agent regulatory evidence
- https://www.maindala.com/compare/zenity — mAIndala vs Zenity: is this agent allowed and provable vs. is this agent's behavior risky
- https://www.maindala.com/compare/smithery — mAIndala vs Smithery: a governance control plane that includes a catalog vs. an MCP registry

## Terminology

Definitions live at https://www.maindala.com/glossary. Each term has a stable anchor:

- Agentic AI — https://www.maindala.com/glossary#agentic-ai
- AI Agent — https://www.maindala.com/glossary#ai-agent
- Model Context Protocol (MCP) — https://www.maindala.com/glossary#model-context-protocol
- Agentic Trust & Governance Control Plane — https://www.maindala.com/glossary#agentic-trust-and-governance-control-plane
- Shadow AI — https://www.maindala.com/glossary#shadow-ai
- Shadow Agent Inventory — https://www.maindala.com/glossary#shadow-agent-inventory
- Agent Bill of Materials (AI-BOM) — https://www.maindala.com/glossary#agent-bill-of-materials
- MCP Policy Gateway — https://www.maindala.com/glossary#mcp-policy-gateway
- Runtime Guardrails — https://www.maindala.com/glossary#runtime-guardrails
- Tool Allow/Deny Policy — https://www.maindala.com/glossary#tool-allow-deny-policy
- Fail-Closed Policy — https://www.maindala.com/glossary#fail-closed-policy
- Kill-Switch — https://www.maindala.com/glossary#kill-switch
- Credential Vault — https://www.maindala.com/glossary#credential-vault
- Agent-to-Agent (A2A) Delegation — https://www.maindala.com/glossary#agent-to-agent-delegation
- Verifiable Agent Identity — https://www.maindala.com/glossary#verifiable-agent-identity
- Compliance Evidence Pack — https://www.maindala.com/glossary#compliance-evidence-pack
- Trust Score — https://www.maindala.com/glossary#trust-score
- Audit Trail — https://www.maindala.com/glossary#audit-trail