M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
C

ChatOllama

Community
Not a standard server
Open-source AI chatbot platform with multi-model support, knowledge bases, realtime voice chat, and MCP integration.

This is not a typical connect-and-use MCP server; read the source repository before using it.

How to install ↓ Compare GitHub source repository ↗
Category
Other #200 of 230
Stars
★ 3.5k Very popular
Transport
stdio (local process) · SSE · Streamable HTTP
Runtime
Node.js 18+ · Docker
Credentials
Optional API key
License
Other / unspecified
Last commit
38FMRS · D

ChatOllama is a feature-rich AI chat platform with MCP integration as a highlight, supporting multiple transports and flexible management permissions. Suitable for developers needing multi-model, knowledge bases, and tool integration, but requires careful security configuration and database migration.

Strongest · Maintenance 12/20 Weakest · Reliability 2/20

Reliability
2/20
Security and permissions
6/20
Maintenance
12/20
Documentation
10/20
Setup experience
8/20
Why each score
Reliability 2/20
The repository is a Nuxt 3 chatbot platform that acts as an MCP client, not an MCP server: the manifest block shows "[not cached]", the README never claims this project exposes an MCP server endpoint (no stdio/SSE/HTTP server entry, no tool list, no handshake description), and all MCP sections explain how to connect external MCP servers. With no server code, tests, or CI evidence in the supplied material, the init + tool-list handshake cannot be verified at all; the static calibration caps this dimension at 12 even with evidence, and here there is essentially none. The recent SQLite→PostgreSQL migration also indicates churn in core storage. Deducted for: not being an MCP server, no tests/CI evidence, no verifiable handshake.
Security and permissions 6/20
The project manages many commercial-model API keys via .env and offers an ACL for MCP management, which is positive. However, the default (and even the "recommended production" line) is ACL_ENABLED=false; per the permission table, unauthenticated and regular users get full MCP management, including adding STDIO servers that execute arbitrary commands — effectively an anonymous, arbitrary-command launch surface on default deployments, a clear overreach. Credential storage details, least-privilege design, and a security-response channel are not documented. No malware, credential theft, or exfiltration evidence was found, so no red-line trigger, but the default open command-configuration surface is a major deduction.
Maintenance 12/20
Recent feature announcements (2025-08-25 ACL, 2025-08-19 Deep Agents, 2025-08-14 PostgreSQL migration) indicate an active project with frequent commits; the repository is not archived and the README points to a Discord community. Deductions: the metadata license is NOASSERTION despite the README claiming MIT, and the LICENSE content cannot be verified from the supplied material; 245 open issues show no visible triage or response evidence; no dependency-update policy or security-response channel is documented.
Documentation 10/20
The README is extensive for the chatbot platform: quick start, Docker and dev setup, environment variables, feature flags, MCP configuration, MCP permission troubleshooting, and user/role management, with well-organized MCP sections. However, documentation of this project as an MCP server is entirely absent (no tool schemas, parameters, or limits), and there are internal contradictions: the env example still shows DATABASE_URL=file:../../chatollama.sqlite while the notice says PostgreSQL is now primary, and the "Data Storage" section still references SQLite paths after announcing the migration. The docs only describe consuming other MCP servers and never disclose the key limitation that this app itself exposes no MCP endpoint to clients.
Setup experience 8/20
The Docker quick start is a single `docker compose up` after downloading docker-compose.yaml, and MCP servers are configured through a documented Settings UI supporting STDIO, SSE, and Streamable HTTP with a uvx example. Deductions: there is no client-connectable endpoint example because this is not an MCP server; development mode requires many external services (Node 18+/pnpm, PostgreSQL, Ollama, ChromaDB/Milvus, Redis); and there is no CI or test evidence that the compose stack actually starts. The static calibration caps this dimension at 15 without execution evidence; the large dependency count and contradictory documentation lower it further.

Static review · not runListed 2026-08-07

Read the FMRS scoring method →

Fit and risk

What it can accessReads local filesWrites / deletes local filesUses the networkConnects to a database

Best for

  • Developers needing multiple language models (OpenAI, Anthropic, Ollama, etc.)
  • Teams requiring local deployment for data privacy
  • Users wanting knowledge bases and realtime voice interaction
  • Developers looking to extend AI capabilities via MCP

Not for

  • Commercial users needing official support
  • Production environments requiring strict MCP management permissions (ACL disabled by default)
  • Users wanting a lightweight, MCP-only solution

Required permissions

  • MCP server management: default all users can manage, with ACL only admins can
  • Tool usage: all users can use configured MCP tools in chat
  • Database access: needs PostgreSQL (dev) or Docker's SQLite/PostgreSQL
  • Filesystem access: STDIO MCP servers may execute local commands

Risks and side effects

  • MCP servers may execute arbitrary commands; configure carefully
  • Improper user role management can lead to security issues
  • Database migration from SQLite to PostgreSQL may require backups
  • Dependence on third-party API keys; leakage can cause costs or data risk

Setup

Before you start

Runtime：Node.js 18+ · Docker

DATABASE_URL requiredsecret
OPENAI_API_KEY optionalsecret
ANTHROPIC_API_KEY optionalsecret
GOOGLE_API_KEY optionalsecret
GROQ_API_KEY optionalsecret
MOONSHOT_API_KEY optionalsecret
COHERE_API_KEY optionalsecret
AUTH_SECRET optionalsecret
Other optional settings (17)
ACL_ENABLED optional
PORT optional
HOST optional
VECTOR_STORE optional
CHROMADB_URL optional
MILVUS_URL optional
NUXT_MCP_ENABLED optional
NUXT_KNOWLEDGE_BASE_ENABLED optional
NUXT_REALTIME_CHAT_ENABLED optional
NUXT_MODELS_MANAGEMENT_ENABLED optional
MCP_ENABLED optional
KNOWLEDGE_BASE_ENABLED optional
REALTIME_CHAT_ENABLED optional
MODELS_MANAGEMENT_ENABLED optional
SUPER_ADMIN_NAME optional
NUXT_PUBLIC_MODEL_PROXY_ENABLED optional
NUXT_MODEL_PROXY_URL optional

See README: Quick start via Docker Compose (docker compose up), or development setup (clone repo, install pnpm, configure .env, create PostgreSQL database, run migrations). For MCP servers, configure via Settings UI or migrate from .mcp-servers.json.

Check that it works

Start the app and open http://localhost:3000, go to Settings → MCP, add a server (e.g. uvx mcp-server-filesystem), enable it, then ask the AI in chat to read a file — if the tool is invoked, the setup works.

Troubleshooting

  1. MCP feature not working: check MCP_ENABLED or NUXT_MCP_ENABLED flag
  2. Cannot access MCP settings after enabling ACL: ensure a super admin exists or set SUPER_ADMIN_NAME
  3. MCP tools not working in chat: verify server config and status, ensure correct transport
  4. Permission changes not taking effect: log out and log in again, or restart app

Things to try

Once connected, you can ask your AI assistant things like:

  • Use the filesystem MCP server to read the README in my project folder and summarize it
  • List the recent commits in my repository using the git MCP server
  • Search the web for the latest MCP protocol news and summarize it
  • Query my local SQLite database for the users table contents

Use cases

Build a local private AI chatbot
Deploy AI agents with tool access
Create RAG knowledge base Q&A systems
Run a multi-model chat platform

Overview

ChatOllama is an open-source chatbot platform built with Nuxt 3, supporting a wide range of language models including OpenAI, Anthropic, Google Gemini, Groq, Moonshot, Ollama, and OpenAI-compatible APIs. It offers multi-modal chat, RAG knowledge bases, realtime voice conversations with Gemini 2.0 Flash, Model Context Protocol (MCP) integration, vector database support (Chroma/Milvus), and Docker deployment. MCP integration supports STDIO, SSE, and Streamable HTTP transports, with a user-friendly interface for managing MCP servers. The platform also includes user management, ACL-based access control, and feature flags for enabling/disabling major areas.

Similar servers

BioMCP 75 · B

One binary. One grammar. Evidence from the biomedical sources you already trust.

★ 654 · Tools 11 Compare with this →
Compare with other MCP servers

Source revision 96c3ce50c3ba Data synced 2026-10-12 Read the FMRS scoring method