mcpField
G
GeniusFlow Evidence

GeniusFlow Evidence

Remote
geniusflow-federation.vercel.appBusiness & Marketing0Updated 2d ago

About GeniusFlow Evidence

GeniusFlow Evidence is an API and MCP service for claims about real-world entities, with source re-checks and citation packages. Its public CIRCLE example uses literal string verification rather than semantic fact-checking.

Key features

  1. 01Retrieve entity cards and published claims
  2. 02Re-fetch sources to check strings
  3. 03Admit or withhold evidence packages
  4. 04Create citation packets for claims
  5. 05Keep missing data marked unknown

Use cases

  1. 01Re-check whether a claim's source still matches
  2. 02Organize evidence admission decisions
  3. 03Prepare citations for published claims

MCP support

The remote MCP server exposes tools for status, entity dossiers, source verification, evidence packaging, and citations. MCP clients can connect to its public endpoint without an account.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://geniusflow-federation.vercel.app/api/mcp
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

Tools

9

This tool takes no parameters.

Details

From the project README

GeniusFlow

Evidence is the product. Federation and MCP are the interfaces. RWA / CIRCLE is the proving ground. Research is the laboratory.

Pinned claims about real-world entities. Live source re-check. Fail-closed packages. Public MCP.

USE NOW      Evidence API + MCP     this README
RESEARCH     labeled experiments    not a product feature
ROADMAP      registry / reliance / settlement    future

Try Federation · Connect MCP · CIRCLE example · What this does not claim


Use GeniusFlow

Evidence-backed claims that can be independently re-checked.

No account. Live origin:

https://geniusflow-federation.vercel.app
Do thisCall
Open an entity cardGET /api/dossier?entity=CIRCLE
Re-check a pinned stringPOST /api/verify
Admit or withholdPOST /api/package
CitePOST /api/cite with {kind, id, claim}
MCPPOST /api/mcp — registry io.github.Kaydeep0/geniusflow-federation

SLA: best-effort Vercel hobby. No promised uptime.

A missing circulating figure stays UNKNOWN. That is product behavior, not an empty cell to fill in.


How it works

Entity → Claim → Grounding → Verify → Package → Cite
  1. Entity — a specimen card (CIRCLE is the public proving ground).
  2. Claim — a published sentence on that card.
  3. Grounding — source_url + expected substring, taken from the claim. Do not invent expected.
  4. Verify — another machine re-fetches the source and reports whether the substring is still there.
  5. Package — admit or withhold. Invented attestations are withheld.
  6. Cite — a cite packet over a published claim object.

Literal string re-check. Not semantic fact-checking. Not “the world is true.”

Pages (kaydeep0.github.io/eigenstate-research) and Federation are different surfaces. Snapshot identities can differ. Do not treat them as one synchronized view.


CIRCLE example

# 1. Entity card
curl -sS "https://geniusflow-federation.vercel.app/api/dossier?entity=CIRCLE"

# 2. Take grounding from claims[0] — do not invent expected
# 3. Verify, package, cite: see examples/

Copy/paste scripts (curl + python3, no clone of the private engine):

Machine index: llms.txt · OpenAPI: openapi.json


Connect MCP

Remote. No private filesystem path.

{
  "mcpServers": {
    "geniusflow-federation": {
      "url": "https://geniusflow-federation.vercel.app/api/mcp"
    }
  }
}

Official registry: io.github.Kaydeep0/geniusflow-federation

Tools (live tools/list): gf_status gf_dossier gf_report_feed gf_cite gf_verify gf_package gf_darshan gf_return_wire gf_openapi

gf_darshan and gf_return_wire are experimental adapters. Core evidence tools are gf_dossier / gf_verify / gf_package / gf_cite / gf_status.


Research

Experiments testing whether maintained state helps independent AI observers recover position and, separately, allocate future attention.

These are labeled experiments. They are not the product on the first screen.

ExperimentPublic status
H_CHATGPT_POSITION R1Task-specific: maintained multi-artifact package C recovered position better than thinner packages A/B (n=15). Not a memory product. Not a general information-advantage proof.
H_CHATGPT_POSITION A1Frozen. 0/25 observers. Waiting on ChatGPT upload apparatus. Not run.
Phase 8Apparatus reachable; prospective experiment not yet started. T0 none. Not a live world-tracking feature. Not an experimental result.

Roadmap

Explicitly future. Not current product claims.

  • RWA registry as something others rely on
  • Shared reliance
  • Settlement

CIRCLE is inspectable today. That does not make a complete RWA registry.


What this does not claim

  • Persistent AI memory product
  • Better than RAG / Graphiti / Mem0 / Letta
  • Complete RWA registry
  • Settlement or reliance
  • Real-time world model
  • Proven information advantage
  • Synchronized websites (Pages and Federation snapshot ids can differ)
  • Production SLA
  • $500T+ tokenized-market figure

Public surfaces

SurfaceURL
Live APIhttps://geniusflow-federation.vercel.app
Human site (secondary)https://kaydeep0.github.io/eigenstate-research/
This repositoryhttps://github.com/Kaydeep0/eigenstate-research
MCPhttps://geniusflow-federation.vercel.app/api/mcp
MCP declarationhttps://geniusflow-federation.vercel.app/server.json

The runtime engine repository is private.

Also on this site

Human research Pages (secondary; not the Evidence API):

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.