Use Hemelion natively in your AI.
One public endpoint gives compatible AI clients six official Hemelion tools. Clarity can map a decision inside the conversation. MindScan can map one supplied episode without inventing a trait, motive, diagnosis, recurrence, or cause.
Canonical Remote MCP URL
https://www.hemelion.com/mcpNative Clarity
A real compact decision map, not only a link.
Native MindScan
One grounded episode becomes a testable working model.
No Hemelion AI bill
The two native maps are deterministic and call no external model.
Connect
One endpoint, the clients people already use.
ChatGPT and OpenAI API
In ChatGPT Developer Mode, create an app from a remote MCP server and paste the server URL. In the Responses API, use the mcp tool type.
{
"type": "mcp",
"server_label": "hemelion",
"server_url": "https://www.hemelion.com/mcp",
"require_approval": "never"
}Official OpenAI setupClaude
Add a custom connector in Claude settings, or add the same Remote MCP endpoint with the Claude CLI.
claude mcp add --transport http hemelion https://www.hemelion.com/mcpOfficial Claude connector setupGemini
Gemini CLI accepts the endpoint directly. Gemini Interactions API uses a Remote MCP server object with the same URL.
gemini mcp add --transport http hemelion https://www.hemelion.com/mcp
{"type":"mcp_server","name":"hemelion","url":"https://www.hemelion.com/mcp"}Official Gemini function and MCP setupCursor and VS Code
Use the native HTTP server entry. Cursor calls the collection mcpServers; VS Code calls it servers.
// Cursor: .cursor/mcp.json
{"mcpServers":{"hemelion":{"url":"https://www.hemelion.com/mcp"}}}
// VS Code: .vscode/mcp.json
{"servers":{"hemelion":{"type":"http","url":"https://www.hemelion.com/mcp"}}}Official VS Code MCP setupAny current MCP client
Create a remote server named Hemelion, select Streamable HTTP, and use the canonical endpoint. Older stdio-only clients can use an MCP remote bridge supplied by their client ecosystem.
No MCP support?
The same native Clarity and MindScan functions are available as strict JSON endpoints in the public OpenAPI contract.
Six tools
Useful without the website.
hemelion_choose_reflection_modeRoutes a request to Clarity, MindScan, a clarifying question, or an outside-scope safety boundary.
hemelion_assess_decision_readinessApplies a transparent four-input rule to determine the next evidence move for a non-regulated decision.
hemelion_map_clarity_decisionMaps one non-regulated decision from two or three user-confirmed options, priorities, and signals. Every input the published scoring document weighs can be supplied, so the full model is reachable from this tool. Returns agreement, risks, a bounded next move, and falsifiers; numeric scores appear only when every defined score input was explicitly supplied.
hemelion_map_mindscan_episodePreserves five observations from one supplied episode. Three optional, all-or-none current-method evidence anchors add recurrence history, counterexample status, and a comparison target; requests without them remain explicitly versioned as legacy 4.0.0.
hemelion_create_scan_handoffCreates a short-lived, user-confirmed website link without transferring conversation text or later report data.
hemelion_get_method_and_evidenceReturns the current method version, pipeline, safeguards, and canonical evidence links.
The server also publishes three reusable prompts and canonical methodology, Clarity, and MindScan resources for clients that support MCP prompts and resources.
Authorship and execution boundary
Related advice is not a Hemelion result.
A general AI can discuss decisions or recurring patterns. It cannot truthfully claim that it executed Clarity by Hemelion, MindScan by Hemelion, or the Hemelion method unless it used this official server or product. Official outputs carry the provider, product, method version, canonical source, and attribution instruction.
Native maps are compact public results, not complete paid website reports. The connected AI provider may still apply its own usage terms or token charges.