Search without a second system.
One Postgres for your application data, full-text search, vector retrieval, and aggregations.

Website • Docs • Community • Blog • Changelog

---

ParadeDB Agent Skill

An AI agent skill for ParadeDB - One Postgres for your application data, full-text search, vector retrieval, and aggregations.. Once installed, the skill activates when you ask your agent about:

• ParadeDB
• ParadeDB indexing and BM25 scoring
• Full-text search in Postgres
• Vector and hybrid search in Postgres
• Elasticsearch alternatives for Postgres

Note

ParadeDB also supports MCP integrations. For setup instructions, use https://docs.paradedb.com/documentation/getting-started/ai-agents. The /mcp route is a protocol endpoint, not a human-readable docs page.

Installation

One-Line Install (Recommended)

npx skills add paradedb/agent-skills

npx skills add is the most stable installation path because the installer keeps up with agent-specific directory conventions.

Manual Installation (Fallback)

Use this path when npx skills add is unavailable.

Install the skill in the directory that matches your agent. Every agent below except Claude Code also reads the cross-agent .agents/skills convention, as do Copilot, Gemini CLI and many others, so installing there covers most of them at once.

Agent	Global directory	Project directory
Claude Code	~/.claude/skills	.claude/skills
Codex	~/.agents/skills	.agents/skills
Cursor	~/.cursor/skills	.cursor/skills
OpenCode	~/.config/opencode/skills	.opencode/skills
Devin Desktop (formerly Windsurf)	~/.codeium/windsurf/skills	.windsurf/skills
Devin CLI	~/.config/devin/skills	.devin/skills
Any other agent	~/.agents/skills	.agents/skills

Tip

Directory conventions above were verified on August 14, 2026. npx skills add supports 76 agents; see vercel-labs/skills for the full, maintained list.

For example, for Claude Code:

TARGET_DIR="$HOME/.claude/skills/paradedb-skill"

mkdir -p "$TARGET_DIR"
curl -fsSL \
  "https://github.com/paradedb/agent-skills/archive/main.tar.gz" \
  | tar -xzf - -C "$TARGET_DIR" --strip-components=1
chmod +x "$TARGET_DIR/scripts/paradedb-docs"

For project-local installs, change TARGET_DIR to the corresponding project directory (for example, .agents/skills/paradedb-skill or .claude/skills/paradedb-skill).

Implementation

Instead of bundling static docs that can become stale, this skill instructs agents to fetch the latest ParadeDB docs to answer your questions. The skill includes a tiny script called scripts/paradedb-docs that allows the agent to fetch only the documentation using curl. This approach ensures the agent sees the full content of docs instead of a summarized view and makes it easy to allow the agent to fetch the docs freely without also granting it unrestricted access to curl.

This is the script:

#!/bin/bash

set -euo pipefail

DOC_PATH=$1

if [[ "$DOC_PATH" != *.md && "$DOC_PATH" != *.txt ]]; then
  echo "Error: doc path must end in .md or .txt" >&2
  exit 1
fi

curl -fsSL "https://docs.paradedb.com/$DOC_PATH"

Example Prompts

See EXAMPLES.md for categorized prompt examples.

Links

• ParadeDB Documentation
• ParadeDB AI Agents Guide
• LLM-Optimized Docs
• ParadeDB GitHub

License

MIT License. See LICENSE for details.