Skip to main content
Coding agents work best with three primitives: docs search, live API tools, and skills (procedural cheat sheets). Praxicraft Assess ships all three.

Pick a path

Agent Plugin (recommended)

One install → Knowledge MCP + API MCP + eight skills

API MCP only

Call Assess tools from Cursor / Claude / Codex

Knowledge MCP only

Search docs — no API credentials

Skills CLI

npx skills add without the full plugin

What each piece does

Knowledge MCP cannot invite candidates. API MCP does not replace reading docs for payload shapes — use both.

Agent Plugin

Source: praxicraft-assess-agent-plugin (Agent Plugins 1.0). Until the public repo is live, clone or symlink from your local checkout.

Cursor

Restart Cursor. Skills load from skills/; MCP servers from .mcp.json.

Claude Code

Codex CLI

Then /plugins → install praxicraft-assess.

Skills included

Try this prompt

Knowledge MCP

Connect any MCP client (no credentials):
Tools search the Assess docs site and read page content. Discovery card: https://docs.praxicraft.com/.well-known/mcp/server-card.json.

Skills

Install Mintlify-hosted skills (and the suite after deploy):
Or install from the Agent Plugin repo:
Also available: llms.txt and .md on every page — see Using docs with LLMs.

API MCP

Full setup (Cursor, Claude Code, Claude Desktop, stdio): MCP (Coding Agents). Hosted OAuth is always live. For Test mode, use stdio with a ct_test_ key via @praxicraft/assess-mcp.

Security

  • Prefer hosted OAuth over pasting live keys into agent config
  • Develop with ct_test_ keys
  • Never commit ct_live_, ct_test_, or whsec_ secrets
  • Review agent-generated handlers for raw-body HMAC verification

Next

MCP · Using docs with LLMs · Webhooks · Integrations · Automations
Last modified on August 24, 2026