Use CVAI’s pro tools inside Claude, ChatGPT, or Grok
Ask your assistant to audit a memo or map a market. CVAI does the work and sends the full report back into the chat. Members sign in once: no keys, no setup.
What you can do
All four pro tools, with the same models and pricing as the website. You talk, the assistant starts the tool, and the report comes back into the chat.
Epistemic Quality Audit
Pulls every consequential claim out of a document, scores it, and flags what is shaky.
Competitive Landscape
Starts from a website, finds the real competitors, and compares offers with evidence.
Value Equation Planning
Generates three distinct offer variants, scored on the value dimensions, with a recommendation.
Pyramid of Evolution
Places an offer on the economic evolution pyramid and points to the next step up.
Connect in 2 minutes
Pick your assistant. There is only one address to remember.
CVAI server address
https://communityvision.ai/api/mcpOn claude.ai and in Claude Desktop, CVAI is added as a connector. You sign in once with your CVAI account.
Add the connector
In Claude, open Settings → Connectors → Add custom connector, then paste this address.
https://communityvision.ai/api/mcpSign in to CVAI
Claude opens the CVAI sign-in page. Sign in, or create an account and pick a plan right there, then authorize Claude.
Ask for an analysis
Paste a document or a link and ask, for example: “Audit the claims in this memo.”
For developers and API users
Everything above also works with an API key, from a script or an agent.
API key
cvai_ak_ keys keep working for scripts, agents, and clients without OAuth. Create one in Dashboard → Developer.
# https://communityvision.ai/dashboard/developer
Authorization: Bearer cvai_ak_your_keyClaude Code, shared with your team (.mcp.json)
At the repo root. The key comes from an environment variable.
{
"mcpServers": {
"cvai": {
"type": "http",
"url": "https://communityvision.ai/api/mcp",
"headers": { "Authorization": "Bearer ${CVAI_API_KEY}" }
}
}
}Codex CLI
In ~/.codex/config.toml.
[mcp_servers.cvai]
url = "https://communityvision.ai/api/mcp"
bearer_token_env_var = "CVAI_API_KEY"OpenAI Responses API
Declare CVAI as a tool of type mcp.
curl https://api.openai.com/v1/responses \
-H "Authorization: Bearer $OPENAI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-5",
"tools": [{
"type": "mcp",
"server_label": "cvai",
"server_url": "https://communityvision.ai/api/mcp",
"headers": { "Authorization": "Bearer cvai_ak_your_key" },
"require_approval": "never"
}],
"input": "Check the balance on my CVAI account."
}'xAI API (Grok)
Same idea on xAI. Their remote MCP field names are still evolving, so check their guide for the exact shape.
curl https://api.x.ai/v1/responses \
-H "Authorization: Bearer $XAI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "grok-4",
"tools": [{
"type": "mcp",
"server_label": "cvai",
"server_url": "https://communityvision.ai/api/mcp",
"authorization": "Bearer cvai_ak_your_key"
}],
"input": "Is CVAI run 1234 finished?"
}'Stay in the loop while it runs
A full audit takes two to five minutes. Your assistant does not wait in silence: it follows each stage, counts the claims processed, and tells you how long is left.
- Reading the document
- Extracting claims
- Scoring each claim12 of 18 claims scored
- Aggregating domains
- Writing the report
- Checking the result
What your assistant says
- Audit started. The document is being read; expect about two minutes.
- Scoring in progress: 12 of 18 claims done, most of them solid so far.
- Report ready. Here are the top three risks and the link to the full report.
Membership
Already a member?
Connect and go. Your plan, your included usage, and your reports are the same as on the website.
Not yet a member?
Connect anyway: the sign-in screen lets you create an account and pick a plan without leaving your assistant.
Your first Epistemic Quality Audit is free, with a results preview.
Plans
- $25 of usage included monthly
- 20% off all usage
- $250 of usage included annually
- 20% off all usage
- No monthly subscription
- Full price (no discount)
Your spend is protected
Same price as the website
A run started from your assistant costs exactly what it costs from the website, plan discount included.
Included usage first
Pro plans include an amount of usage each period. It is used up before anything extra is billed.
No surprises
Past the included usage, billing happens in $50 steps, and only on plans with a card on file.
Everything is visible
Every run shows up in your dashboard with its cost, wherever it was started from.
CVAI in Claude, ChatGPT, and Grok: frequently asked questions
Which AI assistants can use CVAI’s tools?
Any assistant that speaks the Model Context Protocol (MCP): claude.ai and Claude Desktop connectors, Claude Code, ChatGPT connectors, Codex, the OpenAI Responses API, and the xAI API for Grok. Members sign in with their CVAI account; scripts and agents can use an API key instead. The server address is always https://communityvision.ai/api/mcp.
How do I connect CVAI to Claude or Claude Code?
On claude.ai or in Claude Desktop, open Settings → Connectors → Add custom connector, paste https://communityvision.ai/api/mcp, and sign in with your CVAI account when Claude asks. In Claude Code, run claude mcp add --transport http cvai https://communityvision.ai/api/mcp, then type /mcp and pick CVAI to sign in. No API key is needed.
How do I connect CVAI to ChatGPT?
In ChatGPT, turn on developer mode under Settings → Connectors → Advanced, then Settings → Connectors → Create, paste https://communityvision.ai/api/mcp, choose OAuth, and sign in to CVAI. Enable CVAI in the tools menu of a chat and ask for an analysis. The OpenAI Responses API and Codex can also reach CVAI with an API key.
Do I need to be a CVAI member to use the tools from an assistant?
Yes for paid runs: the tools use your CVAI account, plan, and included usage exactly as on the website. Your first Epistemic Quality Audit is free with a results preview. If you are not a member yet, the sign-in screen your assistant opens lets you create an account and pick a plan without leaving the flow; your assistant can also hand you an upgrade link at any time.
What does a run cost when started from an assistant?
The same as from the website. Each run draws from your included usage first, then from your usage balance at your plan’s discount, and every run appears in your dashboard with its cost. Runs never start without a valid account, and per-app daily and monthly limits protect against runaway spend.
Can developers still use an API key with the MCP server?
Yes. cvai_ak_ keys from Dashboard → Developer keep working for scripts, agents, and clients without OAuth, sent as Authorization: Bearer cvai_ak_… on every request. A request without a credential gets a 401 whose WWW-Authenticate header points at the sign-in metadata, which is how assistants start the OAuth flow.