Invisible Back to Dashboard

Invisible Guide

Find NEAR projects with repeat activity, trace a ranking back to its contracts, and give your agent the same data.

Product Walkthrough

Follow a NEAR project from the leaderboard to its on-chain evidence, then see how Codex uses Invisible’s MCP server to answer a research question.

Watch on YouTube

Dashboard

Open the workspace, choose Meteor or another supported NEAR wallet, and sign a message to create your profile. The profile menu shows the NEAR account you used and lets you sign out.

Use Investment for an activity and market comparison, Activity for usage across mapped contracts, or All Verified to see which projects lack enough evidence to qualify. Select a row to update the charts. The magnifier opens its evidence.

Invisible dashboard with the project leaderboard and charts for signer cohorts and scores

Deployment Evidence

Open a project's evidence from the dashboard or leaderboard. Score contributions show how the project performed against peers. Tokenomics lists published allocation and vesting terms. Market figures include their own retrieval date and source.

Rhea Finance evidence showing score contributions and published vesting terms

In Deployments, search by contract address or project. Project pages explain whether an account is a product contract, token, or shared service. Contract pages link to example transactions and show measured cohorts when available. The mainnet block sample refresh is scheduled every twelve hours. Check the saved block time for freshness. Contract transactions are sampled in rotation to stay within the provider limit.

Invisible contract explorer showing the mainnet activity chart and deployment directory

Ecosystem compares changes in tracked contract volume and signer participation. The network gas series covers indexed mainnet activity when a complete series is available.

Forecast

In Leaderboard, choose Forecast and select a project. The flat baseline carries forward the latest signer count. The cohort scenario applies the measured return rate to the latest cohort and carries forward the count seen only in that latest window. A third complete window lets us compare both estimates against what actually happened.

Signer forecast showing observed 28-day windows, a flat baseline, and historical error

How Invisible Measures Activity

We count distinct signing accounts that directly attempted a function call to a mapped NEAR contract. We compare two complete, adjacent 28-day UTC windows. A returning signer appears in both. Calls routed through shared contracts or downstream receipts can fall outside a project's measured activity.

Projects need a verified token, at least 50 signers in each window, 10 in the prior weekly window, 10 returning signers, and activity on at least 14 days in the latest window. A signer is a NEAR account. Shared relayers, automation, and multiple accounts can affect the counts.

35% Adjusted Return Rate
Returning signers divided by the previous cohort plus 50.
30% Returning Signers
Accounts observed in both windows.
25% Adjusted Growth
Net signer change divided by the previous cohort plus 50.
10% Activity Spread
One minus the busiest day's share of the sum of daily signer counts.

Activity scores combine relative percentiles across eligible peers. The investment ranking adds market inputs and requires comparable pool and valuation data.

70%

On-Chain Activity

Usage across the project’s attributed contracts.

20%

Pool Reserves

Largest indexed NEAR pool, compared with ranked peers.

10%

Market Cap / FDV

Reported market cap as a share of fully diluted valuation.

Download the archived pilot report. The archive uses the pilot methodology and its own dated snapshot.

Data Sources

  • NEAR's public BigQuery dataset supplies indexed function-call and signer history. NEAR Public Lakehouse
  • NearBlocks supplies saved mainnet samples, token data, and transaction evidence. NearBlocks
  • DEX Screener supplies pool liquidity and valuation inputs. CoinGecko can supply market rank and valuation when the token can be matched. DEX Screener · CoinGecko
  • Project documentation and code supply attribution and published tokenomics. The ecosystem catalog marks unverified leads separately.

REST API

Send a GET request to the same public data used in the workspace. Read the cutoff and status fields before comparing projects.

1. Read the returning signer ranking

curl -fsS 'https://invisible-gray.vercel.app/api/v1/returning-signers?limit=10'

2. Inspect a project

curl -fsS 'https://invisible-gray.vercel.app/api/v1/projects/rhea'

1. Fetch returning signers

const response = await fetch("https://invisible-gray.vercel.app/api/v1/returning-signers?limit=10");
if (!response.ok) throw new Error(response.statusText);
const ranking = await response.json();

2. Read the project evidence

const evidence = await fetch("https://invisible-gray.vercel.app/api/v1/projects/rhea");
if (!evidence.ok) throw new Error(evidence.statusText);
console.log(await evidence.json());

Endpoints return JSON and support public, read-only access. Use the catalog to resolve a stable project ID before requesting its evidence. Use the market endpoint to check source and observation time separately from the activity cutoff.

/api/v1/projects?q=rhea
Search catalog entries by name, ID, or contract. Use page for pagination.
/api/v1/projects/rhea
Get contract attribution, signer counts, example transactions, and tokenomics for a verified project.
/api/v1/projects/rhea/market
Get separately retrieved market figures for verified tokens.
/api/v1/projects/rhea
Read one project and its evidence.
/api/v1/projects/rhea/market
Read the available pool, valuation and token market evidence.
/api/v1/rankings
Read activity rankings and eligibility.
/api/v1/investment
Read rankings that include liquidity and valuation.
/api/v1/returning-signers?limit=10
Order eligible projects by the count of returning signers.
/api/v1/forecast
Read the next-window scenarios and available historical checks.

MCP

Connect a client using Streamable HTTP. Invisible exposes search_projects, get_project_evidence, and get_rankings.

1. Add the server

claude mcp add --transport http invisible https://invisible-gray.vercel.app/api/mcp

2. Ask in chat

Which NEAR projects brought the most signers back? Show me the evidence.

1. Add the server in .cursor/mcp.json

{
  "mcpServers": {
    "invisible": {
      "url": "https://invisible-gray.vercel.app/api/mcp"
    }
  }
}

2. Ask in Agent chat

Which NEAR projects brought the most signers back? Show me the evidence.

1. Add the server in your terminal

codex mcp add invisible --url https://invisible-gray.vercel.app/api/mcp

2. Ask in a new Codex session

Which NEAR projects brought the most signers back? Show me the evidence.

1. Add to MCP Servers → Manage → View Raw Config

{
  "mcpServers": {
    "invisible": {
      "serverUrl": "https://invisible-gray.vercel.app/api/mcp"
    }
  }
}

2. Reload MCP servers and ask in chat

Which NEAR projects brought the most signers back? Show me the evidence.

  • search_projects resolves a project name, token, or contract to its catalog ID.
  • get_project_evidence returns mapped contracts, source links, and available project evidence.
  • get_rankings reads activity, investment, or returning-signer rankings. Check which cohort and cutoff the result uses.

Ask “Which NEAR projects brought the most signers back? Show me the evidence.”

Example Response in Codex

This example uses the snapshot through November 11, 2026 UTC and ranks eligible NEAR projects with verified tokens.

Codex using Invisible MCP to rank seven eligible NEAR projects by returning signers, with Rhea Finance first at 3,114

Agent-to-Agent

Fetch the agent card, then send an A2A 0.3 JSON-RPC message/send request. The agent accepts returning-signer comparisons, project searches, and verified project IDs. It returns a direct response with research data.

1. Discover the agent

curl -fsS 'https://invisible-gray.vercel.app/.well-known/agent-card.json'

2. Send a research request

curl -fsS 'https://invisible-gray.vercel.app/api/a2a' \
  -H 'Content-Type: application/json' \
  --data '{"jsonrpc":"2.0","id":1,"method":"message/send","params":{"message":{"kind":"message","role":"user","messageId":"invisible-example-1","parts":[{"kind":"text","text":"Which NEAR projects brought the most signers back? Show me the evidence."}]}}}'

1. Discover the agent

const discovery = await fetch("https://invisible-gray.vercel.app/.well-known/agent-card.json");
if (!discovery.ok) throw new Error(discovery.statusText);
const agent = await discovery.json();

2. Send a research request

const response = await fetch(agent.url, {
  method: "POST",
  headers: {"Content-Type": "application/json"},
  body: JSON.stringify({
  "jsonrpc": "2.0",
  "id": 1,
  "method": "message/send",
  "params": {
    "message": {
      "kind": "message",
      "role": "user",
      "messageId": "invisible-example-1",
      "parts": [
        {
          "kind": "text",
          "text": "Which NEAR projects brought the most signers back? Show me the evidence."
        }
      ]
    }
  }
})
});
if (!response.ok) throw new Error(response.statusText);
console.log(await response.json());

Read the agent card to discover the request endpoint and supported skills. Send a JSON-RPC request with a unique message ID and a text prompt, then read the returned research alongside its source links. Read the Agent Card

Agent Skill

Install the skill in your agent's project skills directory. It explains when to use each tool, how to cite evidence, and how to distinguish forecasts from observed activity.

1. Install in your project

mkdir -p .agents/skills/invisible && curl -fsS 'https://invisible-gray.vercel.app/skill/SKILL.md' -o .agents/skills/invisible/SKILL.md

2. Use the skill in a new session

$invisible Which NEAR projects brought the most signers back? Show me the evidence.

1. Install in your project

mkdir -p .claude/skills/invisible && curl -fsS 'https://invisible-gray.vercel.app/skill/SKILL.md' -o .claude/skills/invisible/SKILL.md

2. Use the skill in chat

/invisible Which NEAR projects brought the most signers back? Show me the evidence.

The skill supplies a research workflow. It tells the agent how to find the project, read the ranking inputs, and cite the evidence. The MCP setup above adds the tools to clients that support them. Download SKILL.md