Datris
Open-source data control plane that lets agents acquire, validate, land and query data
- Pricing
- Open source
- Platforms
- self-hosted, CLI, Linux
- Runs
- Remotely or Locally
- Version
- 1.59.0
- Listing updated
- today
Similar MCP servers
About Datris
Datris is an open-source MCP server for analytics and data. It can run locally on your machine or as a hosted remote server.
Datris sits beside your warehouse or lake and gives agents one governed way to work with data, instead of 90 separate integrations. The README describes an operating loop: acquire data with AI-generated taps, validate it with plain-English rules, land it in several destinations in parallel, observe provenance and job state, then explain and repair errors. Pipelines and sync bookmarks live in the platform, so regenerating a script does not lose its place.
Governance is a central theme. Credentials are brokered through a vault, so an agent refers to a secret by name and never holds the key, and agent-written code runs in an isolated container. Each agent gets a scoped API key, an agent policy decides per action whether to run, wait for approval or refuse, and an audit log can be switched on.
You self-host it with Docker, using an AI provider key from Anthropic, OpenAI, Grok, Azure OpenAI or Bedrock. The infrastructure is open source (MinIO, PostgreSQL, MongoDB, Kafka, Vault) under Apache 2.0, and a Homebrew CLI covers ingest, query, search and doctor commands. The full stack wants 8 GB, with a minimal setup of about 3.5 GB.
Install
Run in a terminal
claude mcp add datris --env DATRIS_API_URL='<DATRIS_API_URL>' --env DATRIS_API_KEY='<DATRIS_API_KEY>' -- uvx datris-mcp-serverAdd to claude_desktop_config.json
{
"mcpServers": {
"datris": {
"command": "uvx",
"args": [
"datris-mcp-server"
],
"env": {
"DATRIS_API_URL": "<DATRIS_API_URL>",
"DATRIS_API_KEY": "<DATRIS_API_KEY>"
}
}
}
}Add to ~/.cursor/mcp.json
{
"mcpServers": {
"datris": {
"command": "uvx",
"args": [
"datris-mcp-server"
],
"env": {
"DATRIS_API_URL": "<DATRIS_API_URL>",
"DATRIS_API_KEY": "<DATRIS_API_KEY>"
}
}
}
}Add to .vscode/mcp.json
{
"servers": {
"datris": {
"type": "stdio",
"command": "uvx",
"args": [
"datris-mcp-server"
],
"env": {
"DATRIS_API_URL": "<DATRIS_API_URL>",
"DATRIS_API_KEY": "<DATRIS_API_KEY>"
}
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"datris": {
"command": "uvx",
"args": [
"datris-mcp-server"
],
"env": {
"DATRIS_API_URL": "<DATRIS_API_URL>",
"DATRIS_API_KEY": "<DATRIS_API_KEY>"
}
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"datris": {
"command": "uvx",
"args": [
"datris-mcp-server"
],
"env": {
"DATRIS_API_URL": "<DATRIS_API_URL>",
"DATRIS_API_KEY": "<DATRIS_API_KEY>"
}
}
}
}Using several servers? Build one config file with the MCP config generator.
Configuration
DATRIS_API_URLURL of the Datris server API (default http://localhost:8080)RequiredDATRIS_API_KEYAPI key for the Datris server (only when USE_API_KEYS=true)DATRIS_API_URLURL of the Datris server API (default http://localhost:8080)RequiredDATRIS_API_KEYAPI key for the Datris server (only when USE_API_KEYS=true)
What it does
- One MCP interface
- 90 tools behind a single server.
- Vault-brokered secrets
- Agents reference secrets by name and never hold keys.
- Approval gates
- Agent policies decide whether an action runs, waits for approval or is refused.
- Plain-English validation
- Data quality and transformation rules can be written in natural language.
- Multi-destination landing
- Pipelines write to several stores in parallel with provenance and job state.
- Self-hosted stack
- Runs on a laptop, on-premises or any cloud, under Apache 2.0.
Good for
- Let an agent build a data pull and land it in Postgres
- Validate incoming CSV data with plain-English rules
- Give each agent scoped access with approval for risky actions
- Audit which agent did what to which dataset
- Search landed documents semantically
Pricing
Open source
Apache 2.0, self-hosted; AI provider usage billed separately
Datris FAQ
What does it need to run?
Docker and an AI provider key. The full stack wants about 8 GB of memory, and a minimal install about 3.5 GB.
What licence is it?
Apache 2.0, with open-source infrastructure and no managed service.
Do agents hold my credentials?
No. The README says agents reference secrets by name through a vault.
How do I connect an agent?
Use an MCP client config with the mcp-remote bridge to the bundled SSE server on port 3000, or the PyPI stdio server.
Is Datris free?
Yes. Datris is open source, so you can use and self-host it for free.
How do I add Datris to Claude Code?
Run this command in your terminal: claude mcp add datris --env DATRIS_API_URL='<DATRIS_API_URL>' --env DATRIS_API_KEY='<DATRIS_API_KEY>' -- uvx datris-mcp-server
What are the best alternatives to Datris?
The closest alternatives to Datris are Agent-Native Analytics, Google Ads + Meta Ads + SEO/GEO MCP, dashboard-icons, Tapetide - Indian Stock Market MCP and Midplane.
Where this comes from
- Description: written from the project's README.
- Install steps: from the official MCP Registry entry.

