What your agents have learned, as a map.
A knowledge graph shared across your organisation's agents: the things they know and how those things relate, under a vocabulary they extend as they learn.
What your AI can do with it
- Remember people, projects, systems and how they connect
- Ask how two things are related and get the path between them
- Let every agent build on what another one found
- Start from a template instead of a blank graph
- Search by meaning, not just by name
For example
- You say
- “Who owns the service the billing job depends on?”
- Your AI
- Finds the billing job in the graph, follows its dependency to the service, and reads the owner recorded on it.
- You get
- An answer with its reasoning visible, drawn from what your agents already know.
Its tools
- Create graph
- Find entities
- Explore
- Find paths
- Search
- Create graph
- Begin from a template
- Find entities
- Look things up
- Explore
- Walk a neighbourhood
- Find paths
- How two things connect
- Search
- Search by meaning
Good to know
- Free trial, no card.
- Every price is visible before your AI enables anything.
- Money moves only when you approve a payment link.
- Disabling a service stops its tools and deletes nothing.
Experimental means the tools work and are billed as shown, and their names and shapes may still change before the service is generally available.
Connect your AI
Give your AI this address. It does the rest, and asks you when it needs you.
https://staging.stationzero.services/mcpRun this once in a terminal:
claude mcp add --transport http station-zero https://staging.stationzero.services/mcpInstall with one click, or paste this into .cursor/mcp.json:
{ "mcpServers": { "station-zero": { "url": "https://staging.stationzero.services/mcp" } } }Install with one click, or paste this into .vscode/mcp.json:
{ "servers": { "station-zero": { "type": "http", "url": "https://staging.stationzero.services/mcp" } } }Paste this into your project's AGENTS.md or CLAUDE.md, or give it to your AI as an instruction:
## AI services at https://staging.stationzero.services
This project's AI services are reached over MCP at https://staging.stationzero.services/mcp.
- Your MCP client connects there. The first time, it asks your human to authorize it: they sign in, or create an account, and approve it in their browser. You then act for their organisation. If the server is not in your client's configuration, ask your human to add it.
- Start with `help_explore` and no arguments: it says what the services are, which are on for your account, and lists the help pages. `services_describe` says what a service costs; where it offers a free trial, `services_enable` with `"trial": true` starts one at no cost.
- Anything that costs money is your human's decision: `wallet_top_up` returns a payment link to hand to them.
- The REST API takes an API key, not the MCP token. No key yet? POST https://staging.stationzero.services/api/v1/signup with your human's email address as `operatorEmail` (the Authentication help page explains why): the 201 answer carries `apiKey`, shown once — keep it where you keep secrets and never commit it. Every call creates a new organisation of its own, not your human's, and the key is not accepted at https://staging.stationzero.services/mcp.For any client that reads an mcpServers file:
{ "mcpServers": { "station-zero": { "url": "https://staging.stationzero.services/mcp" } } }No MCP client? One POST creates an organisation and returns its API key:
curl -X POST https://staging.stationzero.services/api/v1/signup \
-H 'Content-Type: application/json' \
-d '{"operatorEmail":"you@example.com"}'This page is written for your AI as well: hand it the address and it can read the full instructions itself.
The first time it connects you sign in, or create an account, and approve it.