Give your LLM a sense of place.
One MCP endpoint. Fifteen spatial tools. Works in Claude Desktop, Cursor, Vercel AI SDK, OpenAI Agents, and any MCP-aware runtime.
Where this comes in.
Google Maps Platform doesn't expose MCP.
The dominant geospatial APIs were designed for mobile SDKs and web maps, not tool-calling agents. Footstep is MCP-native, with token-lean responses, explicit units, and opt-in geometry.
Stitching multiple APIs is rate-limit and pricing chaos.
One auth header covers prediction, routing, geocoding, isochrone, place search, and terrain. One billing line. One response shape.
Your agent needs to actually use the tools.
Footstep tool descriptions are written for LLMs. Render envelopes carry style hints so the agent can draw the response without a transformation step. Field names carry units so the model never has to guess.
Predict access and use
Restricted access. Predict is not part of the standard plan. Access is granted by application only, to organisations with appropriate operational expertise. Apply for access.
Decision support only, never a replacement. Predict outputs are probability priors. They must never replace expert human judgment, established response protocols, or any duty-of-care obligation. Outputs may be incorrect, incomplete, or unsuitable for a given scenario. Final decisions sit with qualified human operators.
Footstep ships a Streamable HTTP MCP server with fifteen geospatial tools. Plug it into Claude Desktop, Cursor, Vercel AI SDK, OpenAI Agents, LangChain, or any other MCP-aware runtime in a single config block.
The same data is reachable over a plain REST API for backends and apps that don't run inside an agent. Same auth, same response shapes. Use both side by side.
What you can count on.
- Native Streamable HTTP MCP server hosted at mcp.footstep.ai.
- Works in Claude Desktop, Cursor, Windsurf, Continue, Vercel AI SDK, OpenAI Agents SDK, LangChain, Gemini.
- Render envelopes carry style hints so agents can draw without transformation.
- Token-lean responses. Geometry is opt-in. Shared admin fields hoisted into a context block.
- Deterministic where it matters, stochastic where it should be.
The tools that fit.
What's near here?
Points of interest near a location, by category or by venue name. Sorted by distance.
Where is this place?
Turn an address, landmark, or place name into coordinates. Returns ranked candidates with confidence scores.
Directions from A to B, by name.
Geocode both ends and compute the route in one call. The agent says 'how do I get from Kings Cross to Tower Bridge'. The tool handles the rest.
Where can you reach in 30 minutes?
Reachability polygons by time or distance from any starting point. Walking, cycling, or driving.
Clean up a messy address.
Free-text in, structured out. Fixes typos, expands abbreviations, infers missing fields, and scores confidence.
Driving, walking, and cycling directions.
Point-to-point or multi-stop routes with terrain analytics. Answer 'is this walkable?' without a second call.
Travel times between every pair.
An N-by-M travel-time matrix between sources and destinations. The basis for assignment, dispatch, and ETA quotes.
Where to look first.
A probability heatmap of where a missing person is most likely to be found, built from their behavioural profile, the terrain, and current weather.
Related use cases
Logistics and autonomous fleets
Multi-stop optimisation, OD matrices, and corridor search for the agents and backends that move things.
Travel and itinerary planning
Find places, route between them, fit them into a time budget, and surface what's on the way. The geospatial layer behind agentic trip planners.
Customer-input cleanup
Fix typos, expand abbreviations, validate before geocoding. The reliable pre-step for agents handling user-typed addresses.