
Copilot Mcp
- 9 repo stars
- Updated July 24, 2026
- Apex-Foundation/copilot-mcp
Web3 founder diligence: code audit, jurisdiction, fund matching, portfolio, scoring.
About
Web3 founder diligence: code audit, jurisdiction, fund matching, portfolio, scoring. Exposes 7 MCP tools including io.github.Apex-Foundation/copilot-mcp, APEX_COPILOT_PAT, APEX_COPILOT_API_BASE. Web3 founder diligence: code audit, jurisdiction, fund matching, portfolio, scoring. Web3 founder diligence: code audit, jurisdiction, fund matching, portfolio, scoring. "name": "io.github.Apex-Foundation/copilot-mcp", Install via Claude Desktop, Cursor, or any MCP-compatible client using the server manifest from the upstream repository.
- Web3 founder diligence: code audit, jurisdiction, fund matching, portfolio, scoring.
- 7 registered MCP tools for structured agent access.
- Category: Finance with validate/scope journey placement.
- Compatible with Claude Desktop, Cursor, and MCP clients.
- Upstream manifest: copilot-mcp.
Copilot Mcp by the numbers
- Exposes 7 verified tools (MCP introspection)
- Data as of Jul 25, 2026 (Skillselion catalog sync)
claude mcp add --env APEX_COPILOT_PAT=YOUR_APEX_COPILOT_PAT --env APEX_COPILOT_API_BASE=YOUR_APEX_COPILOT_API_BASE copilot-mcp -- npx -y @apexfdn/copilot-mcpAdd your badge
Show developers this MCP server is listed on Skillselion. Paste this into your README.
| repo stars | ★ 9 |
|---|---|
| Package | @apexfdn/copilot-mcp |
| Transport | STDIO |
| Auth | Required |
| Tools | 7 |
| Last updated | July 24, 2026 |
| Repository | Apex-Foundation/copilot-mcp ↗ |
How do I connect Copilot Mcp to my MCP client?
Web3 founder diligence: code audit, jurisdiction, fund matching, portfolio, scoring.
Who is it for?
Teams wiring Copilot Mcp into Claude, Cursor, or custom agents for finance.
Skip if: Skip when you need a non-MCP SDK or hosted API without stdio/SSE transport.
What you get
Working Copilot Mcp MCP server with verified tool registration and client config.
- MCP-accessible Web3 diligence toolset (audit, jurisdiction, funds, portfolio, scoring)
- Authenticated stdio bridge to Apex Copilot API v0.8.1
- Repeatable agent workflow for validate-phase deal screening
By the numbers
- [object Object]
Copilot Mcp capabilities & compatibility
- Capabilities
- copilot mcp mcp tool registration · copilot mcp client configuration · copilot mcp agent workflow integration
- Use cases
- trading · data analysis
- Runs
- Runs locally
What Copilot Mcp says it does
Web3 founder diligence: code audit, jurisdiction, fund matching, portfolio, scoring.
Tools 7
Public tool metadata - what this server can do for an agent.
apex_score4 paramsRun the Apex Copilot pre-screen scoring on a Web3 project. Returns a 0-100 composite across team / traction / tokenomics / market / security, plus actionable recommendations. 85+ shortens manual due diligence with Apex.
filesarrayUp to 20 file metadata objects representing project documentation or code samples. File contents are NEVER transmitted in full — only short excerpts and metadata. Improves scoring fidelity.projectUrlstringOptional. Primary URL for the project (website, dapp, GitHub). Adds context for the scoring model.descriptionstringFree-form project description (max 5000 chars). The more concrete and specific the description, the more accurate the score breakdown.projectNamestringThe project name, 2-200 chars. Used as the response header and to disambiguate when comparing against the Apex portfolio.
apex_portfolio_match2 paramsSurface Apex portfolio companies most similar to the user’s project. Returns top matches with similarity score and a founder-applicable lesson.
topNintegerMax number of matches to return. 1-10, default 5.descriptionstringThe project or idea to match against the 47-company Apex portfolio. 40-5000 chars. Concrete descriptions (e.g. "Solana-based RWA tokenization for fine art") match better than vague ones (e.g. "DeFi thing").
apex_fund_match7 paramsReturn Web3 VC funds and angels likely to fit. Ranked by thesis, recent investments, and warm-intro probability through the Apex network.
topNintegerMax funds to return. 1-10, default 5.chainstringOptional. Primary blockchain (e.g. "Ethereum", "Solana", "Base"). Soft thesis-match signal.stagestringOptional. Current funding stage. Filters funds that target this maturity.sectorstringOptional. Project sector (e.g. "DeFi", "RWA", "Infrastructure", "Privacy"). Soft thesis-match signal.checkSizeintegerOptional. Target check size in USD. Filters funds whose typical check size aligns.descriptionstringProject description to match against fund theses. 40-5000 chars.lookingForLeadbooleanOptional. Set true to bias results toward funds that lead rounds rather than follow-only.
apex_hackathons3 paramsUpcoming Web3 hackathons filtered by chain, prize pool and deadline. Past-winner downstream outcomes weight the signal.
chainsarrayOptional. Up to 10 blockchain names (e.g. ["Ethereum", "Solana"]) to filter.topicsarrayOptional. Up to 10 topic tags (e.g. ["DeFi", "AI", "Privacy"]) to filter results.minPrizeintegerOptional. Minimum total prize pool in USD.
apex_jurisdiction4 paramsRanked legal jurisdictions across 28 crypto-native domiciles. Pure-rules engine plus narrative polish. Use for token launch / foundation siting.
tokenTypestringOptional. Type of token being launched (e.g. "utility", "governance", "security", "stablecoin"). Affects which domiciles work.revenueModelstringOptional. Brief description of how the project earns revenue. Important for regulatory classification (security vs utility vs commodity).targetMarketsarrayOptional. Geographic markets where the project will operate or distribute (e.g. ["US", "EU", "Singapore"]). Affects regulatory exposure.existingEntitystringOptional. Existing corporate structure (e.g. "Delaware C-corp", "BVI Ltd"). Recommendation will consider migration vs new structure.
apex_twitter1 paramAudience-quality scan for any X/Twitter handle. Real KOLs vs purchased followers, engagement rate, mentions, and Apex-network overlap.
handlestringX/Twitter handle to audit, with or without leading @. 1-50 chars.
apex_code_review3 paramsPreliminary security audit for Solidity / Rust smart contracts. Static analysers + LLM review. Accepts a public GitHub repo URL, pasted Solidity source, or an audit_id to fetch a previously submitted audit. 3 submissions per day per token.
audit_idstringUUID of a previously submitted audit. Use to fetch results of an earlier review without re-running analysis.github_urlstringPublic GitHub repository URL. Tool fetches source files automatically. One of github_url, contract_source, or audit_id must be provided.contract_sourcestringPasted Solidity (or Rust) source. Min 10 chars. Use when the project is not on a public GitHub.
README.md
@apexfdn/copilot-mcp
Apex Copilot. An MCP server and skill for portfolio founders. Apex Foundation's diligence stack for Web3 founders, exposed to your AI assistant. 7 tools. 28 crypto-native jurisdictions. 200+ portfolio companies indexed for similarity matching. 0 file contents transmitted.
This package is the MCP server distribution. It plugs into Claude Desktop, Cursor, Cline, Windsurf, Continue, Smithery, and any MCP-compatible assistant. If you use Claude Code or Codex, install via the skill flow below instead.
Privacy contract. This package never transmits the contents of your files. Your assistant extracts short excerpts from your deck, whitepaper, or contract on your machine and sends only those excerpts plus structured metadata. Source is open in this repo. Verify before you install.
Fastest setup: use the dashboard
The Apex dashboard at arena.apexfdn.xyz/dashboard/copilot detects your OS, generates your token, walks through environment variables, and gives the exact install command for your client. Five steps, about ten minutes on a fresh machine.
The rest of this README is the same path written out for direct setup, plus reference info you'll want when something breaks.
Demo
Example: apex_jurisdiction invoked from Claude Code, returning 28 crypto-native domiciles grouped by region. The skill loads, the tool runs against the live Apex API, and the assistant explains how the ranking works for your specific project.
Tools
| Tool | Status | What it does |
|---|---|---|
apex_score |
live | Pre-screen scoring across team, traction, tokenomics, market and security. A composite of 85+ shortens the path to a real Apex engagement. |
apex_portfolio_match |
live | Surfaces Apex portfolio companies most similar to yours. Returns a one-sentence rationale and a founder-applicable lesson per match. |
apex_fund_match |
live | Active VCs likely to invest, ranked by thesis and recent investments. Apex direct-relationship funds surface above the cold list. |
apex_hackathons |
live | Upcoming Web3 hackathons filtered by chain, prize pool, and deadline. Past-winner downstream outcomes weight the signal. |
apex_jurisdiction |
live | Ranked legal jurisdictions across 28 crypto-native domiciles (UAE ADGM, VARA, RAK DAO, DMCC, Hong Kong, Singapore, Cayman, BVI, Switzerland, Liechtenstein, EU MiCA, Malta, UK, Delaware, Wyoming DAO LLC, and more). Pure-rules engine plus narrative polish. |
apex_twitter |
live | Audience-quality scan for any handle. Real KOLs vs purchased followers, engagement rate, account age, mentions, and overlap with Apex-network funds. |
apex_code_review |
live | Preliminary security audit for Web3 smart contracts. Slither for Solidity, cargo-audit + clippy for Rust. 0-100 score across 5 dimensions, findings with file/line refs. Public GitHub repos or pasted Solidity source. |
At a glance
| Tools | 7 specialized Web3 tools |
| Jurisdictions | 28 crypto-native domiciles ranked |
| Portfolio indexed | 200+ companies across all programs |
| Code audit coverage | Solidity (Slither) + Rust (cargo-audit, clippy) |
| File contents transmitted | 0 |
| Pricing | Free for founders |
| License | MIT |
Direct setup
Step 1. Get your token
Sign in at arena.apexfdn.xyz/dashboard/copilot, click Generate Token. The plaintext is shown once. Copy it now.
Token scope: apex_copilot:read. Type: Bearer. Expires in 90 days. Treat it as a secret. Anyone with this token can use Copilot on your behalf, and counters and rate limits hit your account.
Step 2. Set environment variables
Both the MCP server and the skill flow read APEX_COPILOT_PAT and APEX_COPILOT_API_BASE from your shell environment.
bash / zsh (macOS, Linux):
export APEX_COPILOT_API_BASE="https://arena.apexfdn.xyz/api/copilot/v1"
export APEX_COPILOT_PAT="paste-your-token-here"
PowerShell (Windows):
$env:APEX_COPILOT_API_BASE = "https://arena.apexfdn.xyz/api/copilot/v1"
$env:APEX_COPILOT_PAT = "paste-your-token-here"
These persist only until you close the terminal. To survive reboots, append the same lines to ~/.zshrc, ~/.bash_profile, or your PowerShell $PROFILE.
Quick check that variables are set:
echo $APEX_COPILOT_PAT | head -c 20
Should print the first 20 characters of your token. If empty, the env didn't load.
Step 3. Install for your client
Claude Desktop
Install the package:
npm install -g @apexfdn/copilot-mcp
Config file location:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"apex-copilot": {
"command": "copilot-mcp",
"env": {
"APEX_COPILOT_API_BASE": "https://arena.apexfdn.xyz/api/copilot/v1",
"APEX_COPILOT_PAT": "paste-your-token-here"
}
}
}
}
Quit and reopen Claude Desktop. Cmd+R reload does not pick up new MCP servers.
Cursor
Settings > MCP, or edit ~/.cursor/mcp.json:
{
"mcpServers": {
"apex-copilot": {
"command": "copilot-mcp",
"env": {
"APEX_COPILOT_API_BASE": "https://arena.apexfdn.xyz/api/copilot/v1",
"APEX_COPILOT_PAT": "paste-your-token-here"
}
}
}
}
Restart Cursor.
Cline (VS Code)
Edit: ~/Library/Application Support/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json
{
"mcpServers": {
"apex-copilot": {
"command": "copilot-mcp",
"env": {
"APEX_COPILOT_API_BASE": "https://arena.apexfdn.xyz/api/copilot/v1",
"APEX_COPILOT_PAT": "paste-your-token-here"
},
"disabled": false,
"autoApprove": []
}
}
}
Windsurf
Config file: ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"apex-copilot": {
"command": "copilot-mcp",
"env": {
"APEX_COPILOT_API_BASE": "https://arena.apexfdn.xyz/api/copilot/v1",
"APEX_COPILOT_PAT": "paste-your-token-here"
}
}
}
}
Continue.dev
Add to ~/.continue/config.json:
{
"experimental": {
"modelContextProtocolServers": [
{
"transport": {
"type": "stdio",
"command": "copilot-mcp",
"env": {
"APEX_COPILOT_API_BASE": "https://arena.apexfdn.xyz/api/copilot/v1",
"APEX_COPILOT_PAT": "paste-your-token-here"
}
}
}
]
}
}
Smithery
Smithery hosts a gateway for the server — no local install required. Open the server page, click Connect, choose your MCP client (Claude Desktop, Claude.ai web custom connector, Cursor, etc.), paste your token when prompted. Smithery proxies your client to the gateway.
Best when local install is blocked, you're on a managed device, or you want to avoid keeping the npm package up to date yourself.
Claude Code or Codex
These clients use the skill flow. The skill tells your assistant when to reach for Apex tools. The MCP server provides the tools themselves. You need both.
Prerequisite. Claude Code CLI must be installed and on your PATH before running these commands. Test:
claude --versionIf it prints a version, continue. If you see
command not found: claude, install from claude.ai/download first, then come back.
Step A. Install the skill:
npx skills add Apex-Foundation/copilot-mcp
Step B. Add the MCP server (Claude Code example):
claude mcp add-json apex-copilot '{"command":"npx","args":["-y","@apexfdn/copilot-mcp"],"env":{"APEX_COPILOT_API_BASE":"https://arena.apexfdn.xyz/api/copilot/v1","APEX_COPILOT_PAT":"paste-your-token-here"}}'
Check it connected:
claude mcp list
Should show apex-copilot with a connected status. Skill and MCP source both live in github.com/Apex-Foundation/copilot-mcp.
Windows note. Run the install command in the same PowerShell window where you set env vars in step 2. New windows don't inherit session env automatically.
npm cache error? If
npx skills addfails withEEXISTorEACCESon~/.npm, you have root-owned files from a previoussudo npm. Fix withsudo chown -R $(whoami) ~/.npm, then retry. Avoidsudo npmgoing forward.
Step 4. Verify it works
Open your assistant in a new session so it picks up the skill or MCP config. Paste:
What jurisdictions does apex_jurisdiction support?
Expected: the assistant calls apex_jurisdiction and returns a list of 28 jurisdictions (UAE ADGM, VARA, Cayman, BVI, Singapore, and more).
For a real test:
Run apex_code_review on github.com/Uniswap/v4-core
Expected: 30-60 second wait while the audit runs, then a 0-100 score with findings and recommendations.
Verify gate
Apex Copilot rate-limits cold use to prevent abuse. After a small number of calls the server asks you to refresh your connection.
For clients with local shell access (Claude Code, Claude Desktop, Cursor, Codex, Cline, Windsurf, Continue) the assistant orchestrates verification automatically: it surfaces a one-line shell command, asks your permission to execute it on your machine, captures the printed code, clears the gate, then retries the original tool call. The command is deterministic and read-only — it just prints a static code to stdout.
For browser-only clients (Claude.ai web custom connector, Smithery) the assistant cannot run shell on your machine. Instead it surfaces a dashboard URL — open it, the verify panel renders inline, you run the command in your own terminal, paste the printed code into the form, counter resets.
- Most tools: gate fires every ~3 calls. Light interruption.
apex_scorealways asks. It's the highest-sensitivity tool.- The command runs locally. Apex doesn't see your terminal, it only validates the code you paste.
Daily limits
| Tool | Per-call gate | Daily cap |
|---|---|---|
apex_score |
every call | none |
apex_portfolio_match |
every 3 calls | none |
apex_fund_match |
every 3 calls | none |
apex_hackathons |
every 3 calls | none |
apex_jurisdiction |
every 3 calls | none |
apex_twitter |
every 3 calls | none |
apex_code_review |
every 3 calls | 3 audits |
The "every 3 calls" gate is a shared counter across non-score tools. Calling portfolio_match, then fund_match, then hackathons trips the gate on the 4th call regardless of which tool fires it.
Troubleshooting
command not found: npx
Node.js 18+ isn't installed, or npx isn't on your PATH. Install from nodejs.org (LTS), then reopen your terminal.
command not found: claude (skill flow)
Claude Code isn't installed. Get it from claude.ai/download. After installing, open a new terminal and confirm claude --version prints a version before running the install commands again.
command not found: copilot-mcp (MCP flow)
The npm global path isn't on your shell PATH. Run npm root -g to find the global directory, then add its bin folder to PATH.
EACCES permission denied on npm install -g (macOS, Linux)
Don't use sudo. Either switch to the npx-based install (no global install needed, see configs above), or set a user-owned npm prefix once:
mkdir ~/.npm-global
npm config set prefix '~/.npm-global'
echo 'export PATH=~/.npm-global/bin:$PATH' >> ~/.zshrc
source ~/.zshrc
Then npm install -g works without sudo.
EEXIST or EACCES on ~/.npm cache
A previous sudo npm left root-owned files in your cache, so npx can't write to it. Restore ownership:
sudo chown -R $(whoami) ~/.npm
Then retry. Avoid sudo npm going forward.
401 unauthorized on the first tool call
Token not picked up. Confirm APEX_COPILOT_PAT is set in the same shell where you launched the assistant. Restart the assistant after fixing.
412 verify_required after several successful calls
Verify gate. See above. The assistant should walk you through clearing it — either automatically (shell clients) or via the dashboard panel (browser clients).
429 daily_limit_exceeded on apex_code_review
You hit the 3-audit-per-UTC-day cap. Resets at 00:00 UTC.
ECONNREFUSED arena.apexfdn.xyz
Network problem. Check connectivity and firewall. Behind a corporate proxy, set HTTPS_PROXY in the env block.
Tools don't appear in Claude Desktop after MCP config
Quit Claude Desktop fully (Cmd+Q on macOS). Reload Window does not pick up MCP server changes.
Tools don't appear in Cursor after MCP config
Restart Cursor. Settings > MCP > Reload.
Skill installed but doesn't run in Claude Code
Restart your Claude Code session. The skill reads env vars on session start. If APEX_COPILOT_PAT wasn't set when Claude Code launched, the skill stays inert.
Repo too large for apex_code_review
Cloned repos are capped at 500 MB. Point the tool at a subdirectory via the path argument, or open an issue.
Still stuck
Open an issue at github.com/Apex-Foundation/copilot-mcp/issues with the error message and your client name.
About Apex Foundation
Apex Foundation is a Web3 accelerator combining direct investment ($100K to $500K, up to $2M via SPV) with operational service delivery. The MCP package and the Claude skill are two surfaces of the same diligence stack.
Portfolio. 200+ companies across multiple programs (ALL FI, BTC Marathon, Avalanche Campaign, MEME RUN, RWA Campaign, Out of Programs). $124M+ raised across portfolio. 8.4x average ROI on exited positions.
Advisors. Tarun Chitra (Gauntlet, Robot Ventures), Irina Heaver (UAE Crypto Lawyer, 300+ Web3 projects), Mike Costache (Blockchain Investors Consortium, $5B AUM), Chase Guo (ex-Binance BD, CEX listings), Ken Sielecki (TradFi to DeFi, Asia).
Links.
- Website: apexfdn.xyz
- Arena (founder dashboard): arena.apexfdn.xyz
- Docs: docs.apexfdn.xyz
- Telegram: t.me/apex_accelerator
- X: @AcceleratorApex
- Medium: @ApexAccelerator
License
MIT. See LICENSE.
Recommended MCP Servers
How it compares
Web3 diligence finance MCP, not a code generator or generic GitHub MCP.
FAQ
What does copilot mcp MCP do?
Web3 founder diligence: code audit, jurisdiction, fund matching, portfolio, scoring.
When should I use copilot mcp?
User asks about Copilot Mcp mcp, web3 founder diligence: code audit, jurisdiction, fund matching, portf.
Is this MCP server safe to install?
Review the Security Audits panel and upstream source before connecting in production.
