Agent connectors for India's economy — and your enterprise
The Reserve Bank's data warehouse, national statistics, Union Budgets, parliamentary records, live vehicle registrations, geospatial intelligence — and, next to them, connectors pointed at your own material: internal datasets, vector collections, scheduled web extraction, document redesign, an expert visualiser. Every one is something any AI agent can call.
Try them live in the browser, wire them into your own stack, or mount a few and turn them into an app with instructions of your own — run it here, share it as a link, or drop it into your site as an embedded widget.
Pick a client, choose the MCPs to connect, copy one config — then test the same tools live in each playground.
Submit a client
Anthropic · Proprietary
Anthropic's terminal agent. First-class remote MCP with in-CLI OAuth and dynamic client registration — the fastest path from a shell to these data sources.
OpenAI · Apache-2.0
OpenAI's open-source terminal agent. Remote MCP servers are declared in config.toml, with codex mcp login for OAuth sign-in.
Google · Apache-2.0
Google's open-source terminal agent. Remote MCP over httpUrl with automatic OAuth discovery and dynamic client registration on 401.
Anthropic · Proprietary
Build your own harness on the machinery behind Claude Code — remote MCP servers are one option in the query config.
OpenAI · MIT
OpenAI's open-source agent framework — remote MCP servers attach to any agent as hosted tool collections.
Listed here are the first-party clients from Anthropic, OpenAI and Google — the ones whose auth path we re-verify against the gateway on every release. Any other MCP client still works: these are remote MCP servers over HTTP with OAuth 2.1 or a bearer key, so hosted clients — claude.ai, Claude Desktop, ChatGPT, Cursor, VS Code — connect with just the gateway URL, and each MCP's Docs page has the walkthrough.