Developers

Connect an agent to Nous

One URL is an MCP server: real-world patterns for positions and expertise, from groups of 50 or more, and skills written from them. No individual records.

https://www.taicheck.com/api/v1/intelligence

1. Get a key

Sign in, open the dashboard → Connect and create a key. It is shown once. Keep it in an environment variable, and never paste it into a chat with an AI. Plans are monthly pools of credits (Starter, Builder, Pro), or a free month of Starter for 10 rounds on Contribute within 7 days (short scenarios or pairs to rate); see Account → Plan & billing.

export NOUS_API_KEY=tai_live_…

Create a key

2. Check the connection

tai doctor checks the key, the server, the MCP handshake, the tools, a first call and your usage, with the fix for anything that fails.

curl -fsSLo tai.mjs https://www.taicheck.com/tai.mjs
node tai.mjs doctor

3. Add Nous to your agent

Claude Code

In a terminal:

claude mcp add --transport http nous https://www.taicheck.com/api/v1/intelligence \
  --header "Authorization: Bearer $NOUS_API_KEY"

Claude Desktop

Settings → Developer → Edit config (it connects through mcp-remote):

{"mcpServers":{"nous":{"command":"npx","args":["-y","mcp-remote","https://www.taicheck.com/api/v1/intelligence",
  "--header","Authorization:${TAI_AUTH}"],"env":{"TAI_AUTH":"Bearer tai_live_…"}}}}

Cursor

~/.cursor/mcp.json:

{"mcpServers":{"nous":{"url":"https://www.taicheck.com/api/v1/intelligence",
  "headers":{"Authorization":"Bearer ${env:NOUS_API_KEY}"}}}}

VS Code

.vscode/mcp.json (asks for the key once):

{"inputs":[{"type":"promptString","id":"nous-key","description":"Nous API key","password":true}],
 "servers":{"tai":{"type":"http","url":"https://www.taicheck.com/api/v1/intelligence",
  "headers":{"Authorization":"Bearer ${input:nous-key}"}}}}

Gemini CLI

~/.gemini/settings.json:

{"mcpServers":{"nous":{"httpUrl":"https://www.taicheck.com/api/v1/intelligence",
  "headers":{"Authorization":"Bearer $NOUS_API_KEY"}}}}

ChatGPT (connector)

Settings → Apps & Connectors → Advanced → Developer mode on; then Create, paste the URL and choose OAuth. ChatGPT asks you to sign in to Nous once: no key to paste.

https://www.taicheck.com/api/v1/intelligence

ChatGPT (Custom GPT)

In the GPT editor: Configure → Actions → Import from URL; Authentication: API key, Bearer, your tai_live_ key.

https://www.taicheck.com/api/v1/openapi.json

OpenAI Codex

~/.codex/config.toml:

[mcp_servers.tai]
url = "https://www.taicheck.com/api/v1/intelligence"
bearer_token_env_var = "NOUS_API_KEY"

OpenAI Responses API

As a remote MCP tool in your request (the key from your server's environment):

tools: [{type: "mcp", server_label: "tai", server_url: "https://www.taicheck.com/api/v1/intelligence",
  headers: {Authorization: `Bearer ${process.env.NOUS_API_KEY}`}, require_approval: "never"}]

The tai CLI

No MCP client needed: it calls the same tools.

curl -fsSLo tai.mjs https://www.taicheck.com/tai.mjs
node tai.mjs discover "Write a usability findings report for a checkout redesign"
node tai.mjs skill position/design/senior --save

Custom connectors on claude.ai need OAuth, which Nous does not offer yet.

What your agent gets

tai_discover_intelligence
Start here: describe the task in words.
tai_learning_path
The positions and expertise a task needs, in order.
tai_curriculum
How an agent learns a role, level by level: knowledge to load, tools, skills, practice with checks.
tai_get_skill
A SKILL.md for a position or expertise, written from Nous data.
tai_get_skill_context
The report, workflow and rules for one task.
tai_plan_run
Each step's tool, how to call it and what it must produce; tai_report_step checks each step, tai_check_output (with structure) the file.
tai_check_updates
Whether what you fetched has changed, and what changed.
tai_skill_feedback
How a skill worked for your task. Not billed.

Figures describe groups and are context, never requirements for a person. Patterns are worldwide.