Market intelligence agent manages its own data pipelines via MCP
A developer built a market intelligence agent on top of Datris, an agent-native data platform, that uses MCP to autonomously create pipelines, ingest live financial data, and answer queries — with no pre-loaded datasets.
Score breakdown
Developers building agentic data workflows can study this as a concrete pattern for letting agents manage infrastructure dynamically via MCP, rather than querying static, pre-built datasets.
- 01Built on Datris, an agent-native data platform developed with Claude assisting.
- 02The agent calls `tools/list` at startup to discover MCP tools at runtime — no tool names are hardcoded.
- 03Pipeline lifecycle operations (create, ingest, query) are exposed as MCP tools, making MCP the control plane between agent and data platform.
The author built a market intelligence agent that treats data infrastructure as something an agent actively manages, not just queries. The system is built on Datris, an agent-native data platform developed with Claude assisting, which exposes pipeline lifecycle operations — create, ingest, query — as MCP tools. Rather than hardcoding tool names or pre-loading datasets, the agent calls `tools/list` at startup to discover what capabilities are available at runtime, then decides what pipelines to build and what data to pull based on the user's query.
The front-end is a FastAPI + vanilla JS application (no Node) that connects to the Datris MCP server.
The front-end is a FastAPI + vanilla JS application (no Node) that connects to the Datris MCP server. When a user asks something like "What's the current macro picture?", the agent discovers relevant MCP tools, decides it needs FRED macro data (yields, VIX, CPI, credit spreads) plus equity OHLCV data, creates pipelines for each source, ingests live data from free public APIs including FRED, yfinance, CoinGecko, and SEC EDGAR (10-K/10-Q filings), and returns a cited answer. A live pipeline status panel in the browser is kept current via an SSE stream. Background refresh keeps active pipelines updated automatically.
The key design insight is using MCP as the control plane between the agent and the data platform, rather than building a custom API layer. Because the agent discovers capabilities dynamically, adding a new tool to the MCP server makes it immediately available to the agent with no code changes. The platform is open-sourced under AGPL at `github.com/datris/datris-platform-oss`, with example code for the market macro agent available in the repository.
Key facts
- 01Built on Datris, an agent-native data platform developed with Claude assisting.
- 02The agent calls `tools/list` at startup to discover MCP tools at runtime — no tool names are hardcoded.
- 03Pipeline lifecycle operations (create, ingest, query) are exposed as MCP tools, making MCP the control plane between agent and data platform.
- 04All data sources used are free: FRED, yfinance, CoinGecko, and SEC EDGAR (10-K/10-Q).
- 05The front-end is a FastAPI + vanilla JS app (no Node) with live pipeline status streamed via SSE.
- 06Adding a new tool to the MCP server makes it automatically available to the agent without code changes.
- 07The platform is open-sourced under AGPL at `github.com/datris/datris-platform-oss`.