§7 Case studies

A modular market-update agent for an Indian trading community

A growing trading community wanted timely, factual market updates in its Telegram group without a moderator writing every one. We built a modular agent package on Hermes Agent with read-only MCP servers for market data and exchange filings. It posts plain-language updates three times every trading day, and members can pull the same package to run and adapt in their own workflows.

3 a dayfactual updates on trading days
Read-onlyno trades, no recommendations
Modularmembers pull, run and adapt it
Partner
Confidential, at their request
Industry
Trading education
Region
India
Built with
Hermes Agent · Python · Model Context Protocol · Angel One SmartAPI · NSE and BSE public feeds · Telegram
Diagram: read-only market-data and filings MCP modules feed a Hermes agent that posts opening, mid-day and closing updates.

The challenge

Every trading day, members wanted the same three things: where the indices stand, what moved in futures and options, and which NSE and BSE announcements came out. A moderator pieced that together by hand from price screens, exchange filings and news sites before each update. It took hours, and on the busiest days the update arrived late.

What we built

A reporter, not an advisor. The agent runs on Hermes Agent and writes from the day's data. Its instructions are explicit: state what the data shows, never what anyone should do. It gives no buy or sell calls, no targets and no tips.

Two read-only data connections. We built two MCP servers for the agent. One reads market data, index levels and the day's biggest futures and options movers through a broker's market-data API. The other gathers NSE and BSE disclosures from their public feeds, Indian and global market news, the exchange holiday calendar and the overnight US close. Neither can place a trade.

A schedule that knows the market. Three updates on trading days and two news briefings on market holidays, following the exchange's published calendar.

Answers on demand. Members type /market or /news in Telegram and get a fresh summary in seconds.

A package members can make their own

The whole agent ships as a modular Hermes and MCP package that any member can pull and run with their own credentials.

  • Plug it into your own workflow: post to your own group, run it on a schedule that suits you, or call it on demand.
  • Change what it writes: the post format and sections live in one editable skill file.
  • Swap or add data sources: each MCP server is a separate module with its own tests, so one can be replaced without touching the rest.
  • Safe to set up: every secret is an environment variable, a setup script installs each server, and a smoke test checks every data source before the first post.

Built to be trusted

  • Read-only by design. The agent can look at the market but can never act in it.
  • Recovers on its own. Broker sessions expire every night; the connection signs back in before that happens, with no restart and no one on call.
  • Shows its sources. Every news item is marked as an official exchange filing or a media report, and a feed that fails says so instead of going quiet.
  • Tested on every change. Automated tests run on each update, and a live check exercises every data source before the agent goes back to work.

The outcome

  • 3 factual updates every trading day, written and published with no one at the keyboard.
  • Hours of daily research moved from a moderator to the agent.
  • A package the community owns: members run, modify and extend it for their own use.
  • Runs unattended on the community's own server, fully documented.

Important: the agent publishes factual market summaries for information only. It does not provide investment advice, recommendations or research, and it does not place trades. Each deployment uses its operator's own API credentials and is responsible for following its data providers' terms. This project is not affiliated with or endorsed by NSE, BSE or Angel One.


An agent with narrow, read-only access to your systems, posting where your people already work: the same pattern fits sales digests, operations alerts and daily support summaries.

Tell us what you want to build.