Browse documentation
Build with Drishti MCP and Coding Agents
Drishti hosts a public Model Context Protocol (MCP) server for Indian equities intelligence and Drishti developer documentation.
Server URL (streamable HTTP): https://mcp.drishti.manasija.in
What is MCP?
Model Context Protocol (MCP) is an open standard for connecting an AI assistant to external tool servers. The assistant sees a list of tools with defined inputs and outputs, then calls them during a task.
That is different from a generic web fetch:
| Generic web access | Drishti MCP |
|---|---|
| The agent browses pages or URLs ad hoc | The agent calls named tools like search_docs, get_announcements, or search_concalls |
| Page HTML must be parsed and interpreted | Responses come back in a structured shape the tools define |
| No built-in auth to your account | Your Drishti API key authenticates requests to your plan |
| Docs and data are mixed with the open web | You get Drishti documentation and Indian market data through one connector |
So MCP does not replace web access. It gives your coding agent reliable, authenticated access to Drishti tools while you build: read official docs, inspect real API payloads, and pull market context into the same chat where you write code.
In practice, that means you can stay inside Cursor, Codex, Claude Code, or VS Code and ask:
- "What are the query parameters for
GET /v1/announcements?" - "Fetch the latest announcements for RELIANCE and TCS."
- "Scaffold a watchlist digest using the correct Drishti SDK fields."
The agent calls Drishti MCP tools, then uses the returned docs or data in its answer or in the code it generates.
What Drishti MCP provides
- Indian market data tools: filings, earnings, concalls, news, events, prices, sectors, movers, and symbol lookup.
- Documentation tools: search and read Drishti guides, SDK docs, and API reference pages.
- Research routing helpers: prompts and playbooks that steer the agent toward the right tool for a question type.
Drishti MCP is for research, exploration, and assisted development. It does not replace a production runtime. When you ship a scheduled job, dashboard, or alert system, move deterministic fetches to the SDK or WebSocket streams.
Prerequisites
- A Drishti API key from the platform console.
- Node.js 18+ if you use the `drishti-mcp` installer.
- An MCP-compatible coding agent with HTTP transport support.
Quickstart
The fastest path is the drishti-mcp installer. It detects supported clients on your machine and writes the correct MCP config with your API key.
pnpm dlx drishti-mcpRestart your coding agent after setup so it reloads MCP settings.
For non-interactive setup:
DRISHTI_MCP_API_KEY=your-api-key pnpm dlx drishti-mcpThe installer auto-detects these clients:
- Cursor
- VS Code
- Zed
- Codex
- Claude Code
- Antigravity CLI
Client-specific setup
MCP tools for coding
Drishti MCP exposes three tool families. Your agent picks the family based on what you are trying to do.
Documentation tools
Use these while writing integration code so the agent reads official docs instead of inventing endpoints.
| Tool | When to use it |
|---|---|
search_docs | Find a guide, SDK page, or API reference entry by keyword (auth, websocket, pagination, pricing, etc.). |
get_doc | Fetch the full markdown for one docs page after search, or when you already know the path (for example /docs/guides/websockets). |
read_me | Load routing guidance first on open-ended Indian market questions so the agent picks sensible data tools. |
Typical coding flow: search_docs → get_doc → generate code against the documented request/response shape.
Market data tools
Use these when the agent needs live or recent Indian market context while building or debugging a tool.
| Tool | When to use it |
|---|---|
resolve_symbols | Match a company name or phrase to likely NSE/BSE symbols before any symbol-specific fetch. |
get_symbols | Fetch metadata for known tickers or scrip codes. |
get_announcements | List exchange filings and corporate announcements. |
search_announcements | Search filing text by theme or phrase across companies. |
get_announcement_categories | List announcement category values for filters. |
get_news | Fetch recent company or market news. |
get_events | Fetch corporate events such as board meetings and AGMs. |
list_earnings | Browse earnings filings across companies or a watchlist. |
get_earnings_filing | Fetch one company's earnings report for a specific quarter. |
get_concalls | List conference call records. |
get_upcoming_concalls | List scheduled calls that have not happened yet. |
search_concalls | Search concall transcripts for management commentary on a topic. |
get_price_and_volume_since | Fetch historical closing price and volume over a window. |
get_sectoral_data | Fetch sector and company background for symbols. |
get_sector_wise_companies | List peer companies in the same sector or industry. |
get_top_movers | Fetch the day's biggest gainers and losers. |
get_top_traded | Fetch the most actively traded symbols. |
get_52w_data | Fetch 52-week high and low context. |
get_daily_summary | Fetch the market daily summary digest. |
Typical coding flow: resolve_symbols (if needed) → one or more fetch/search tools → shape output in the script or UI you are building.
Prompt and routing tools
Use these when you want more consistent agent behavior across a project or team.
| Tool | When to use it |
|---|---|
get_default_system_prompt | Baseline Drishti research instructions for an agent session. |
get_company_research_prompt | Company deep-dive workflow guidance. |
get_earnings_concall_prompt | Earnings and concall analysis workflow guidance. |
get_market_development_prompt | Market-wide development and news workflow guidance. |
get_subagent_prompt | Prompt pack for delegating a sub-task to another agent. |
MCP also exposes instruction resources (for example drishti://instructions/tool-routing) for clients that read MCP resources. Prefer one instruction source per session to avoid duplicate guidance.
Use cases
Look up Drishti docs while you code
Problem: You are integrating Drishti REST, WebSocket, or SDK APIs and do not want to tab out to the docs site for every parameter, error code, or billing rule.
How MCP helps: The agent calls search_docs and get_doc to pull the official page into context, then writes code against the documented contract.
Example prompt:
Search the Drishti docs for WebSocket symbol limits and authentication, then explain what changes between Starter and Pro.Explore real payloads before you commit to a schema
Problem: You are building a watchlist digest, dashboard, or alert formatter and need to know which fields Drishti actually returns.
How MCP helps: The agent calls market data tools on your symbols, inspects live payloads, and scaffolds types, parsers, or UI columns that match reality.
Example prompt:
Call Drishti MCP for RELIANCE, TCS, and INFY announcements and news from the last 7 days.
Show me the fields we should map in a digest table, then generate a TypeScript interface.Research chat with cited market context
Problem: You want to ask follow-up questions about Indian equities during development without copying filings, news, or transcript excerpts into the chat yourself.
How MCP helps: The agent calls the right data tools per question, keeps symbols and dates from the response, and cites which tool produced each fact.
Example prompt:
What changed for INFY in the last week across announcements, news, and earnings context?
Cite the Drishti tool used for each bullet.Build market event tracking with an agent
Problem: You are creating a script, bot, or internal tool that surfaces "what happened on my watchlist" and you want the first version fast.
How MCP helps: During development, the agent uses MCP to validate event types, test filters, and prototype output. You then harden the runtime path on the SDK or WebSocket layer for production.
Example prompt:
Help me build a watchlist event tracker for RELIANCE, TCS, and INFY.
Use Drishti MCP to confirm payload shapes, then scaffold a CLI that prints announcements, news, events, and earnings with source attribution.
Note where production code should use drishti-sdk instead of MCP.Onboard a team to Drishti APIs
Problem: New developers on your team ask the same integration questions repeatedly.
How MCP helps: A shared AGENTS.md plus Drishti MCP gives every coding agent the same docs and data access pattern, which reduces wrong assumptions about endpoints and field names.
Example prompt:
Explain how Drishti authentication works for REST and WebSocket using the official docs, then show curl and SDK examples.Verify the connection
Docs check
Search the Drishti docs for "pagination" and summarize how list endpoints handle limits and date windows.You should see search_docs and get_doc in the tool trace.
Data check
Use Drishti to check the latest announcements, earnings updates, concall context, and news for RELIANCE and TCS.
Summarize the most important changes and cite which tool each point came from.You should see market data tool calls and real symbols, dates, and headlines in the answer.
MCP vs SDK: when to use which
| Situation | Use MCP | Use SDK or WebSocket |
|---|---|---|
| Exploring docs and response shapes in chat | Yes | No |
| One-off research while building | Yes | Optional |
| Production cron jobs or services | No | Yes |
| Near-live alerts and push delivery | No | WebSocket streams |
| Unit tests and deterministic retries | No | Yes |
Use MCP to learn and prototype. Use the SDK or WebSocket layer to ship.
Tips
- If you do not add guidance to
AGENTS.md, explicitly tell the agent to use Drishti MCP for Indian market or Drishti API questions. - If you run multiple MCP servers, keep server names short and descriptive (for example
drishti) so the agent can pick the right one. - For ambiguous company names, tell the agent to call
resolve_symbolsbefore symbol-specific fetches. - For docs questions, prefer
search_docsthenget_docinstead of asking the model to recall endpoint details. - MCP tool calls consume Drishti credits on data tools. See Pricing for plan behavior.
- Treat your API key as a secret. Do not commit MCP config files that contain the Bearer token.
Security notes
- Only connect to the official endpoint:
https://mcp.drishti.manasija.in. - Rotate your API key from the platform console if a config file or screenshot leaks.
- Do not expose your Drishti API key in browser bundles or public repositories.
Related docs
- Drishti MCP reference: full tool catalog, OAuth flow, and manual install.
- Drishti MCP for chat: wire Drishti tools into a chat backend with any agent runtime.
- Live chat demo: hosted reference app.
- chat-drishti-mcp on GitHub: source code.
- Connect MCP to Claude: web assistant connector workflow.
- How to Set Up WebSockets: near-live delivery for production alerting.
- Authentication: API key setup for REST, WebSocket, and MCP.
- JavaScript / TypeScript SDK and Python SDK: production runtime clients.
- Pricing: REST credits, MCP usage, and plan limits.