Pointmoon for agents

Give an agent a sense of what is happening outside.

Pointmoon is a public, read-only MCP server for current physical and living-world context. Give it a place or coordinate and it returns sourced field truth — weather, light, season, phenology, nearby observations, water, terrain and related signals — or explicit silence when the data cannot support an answer.

One URL. No account. No API key.

Remote MCP

Start here.

https://pointmoon.ai/api/mcp

Add that URL as a Streamable HTTP MCP server in any compatible agent client. Pointmoon exposes two public tools: field_truth for the sourced current-state substrate, and step_outside for the narrower “is this a good moment to go outside?” decision.

Claude Code

One command.

claude mcp add --transport http pointmoon https://pointmoon.ai/api/mcp

Cursor / project config

One small config.

{
  "mcpServers": {
    "pointmoon": {
      "url": "https://pointmoon.ai/api/mcp"
    }
  }
}

No MCP client

Plain HTTP works too.

curl "https://pointmoon.ai/api/moon?audience=facts&place=Boston"

The MCP tool and HTTP endpoint resolve through the same Pointmoon field-truth pipeline. Pointmoon does not write the final sentence for the model. It returns tokens with provenance and confidence, with freshness on the underlying readings.

When an agent should call it

Current world state, not generic knowledge.

1

What is happening here now?

Current physical and living-world conditions at a named place or coordinate.

2

Ground this outdoor decision.

Give a travel, walking, nature or outside recommendation real place-and-time evidence.

3

Do not guess.

If a source is missing, stale or too uncertain, Pointmoon leaves that axis unresolved instead of inventing a condition.

The model speaks. Pointmoon grounds it.

The server is public and read-only today. The open connector and conformance suite are on GitHub.

MCP connectorPointmoon source