Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/mnott/PAIWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/plugins/mnott/pai/gemini-extension)<a href="https://agentmods.dev/plugins/mnott/pai/gemini-extension"><img src="https://agentmods.dev/badge/plugins/mnott/pai/gemini-extension/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/plugins/mnott/pai/gemini-extension"><img src="https://agentmods.dev/badge/plugins/mnott/pai/gemini-extension.svg" alt="Reviewed on agentmods" width="80" height="20"></a>Grade A, and why
@tekmidian/pai scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 16d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
{
"name": "@tekmidian/pai",
"displayName": "PAI Knowledge OS",
"description": "Personal AI Infrastructure — persistent memory and session continuity for AI coding assistants",
"version": "0.7.0",
"author": "Matthias Nott",
"license": "MIT",
"homepage": "https://github.com/mnott/PAI",
"mcpServers": {
"pai": {
"command": "node",
"args": ["dist/daemon-mcp/index.mjs"],
"description": "PAI Knowledge OS — federated memory search, project registry, session management. Provides 10 tools: memory_search, memory_get, memory_outline, project_info, project_list, session_list, registry_search, project_detect, project_health, project_todo."
}
},
"capabilities": {
"mcp": true,
"hooks": false,
"skills": false
},
"setup": {
"prerequisite": "Run 'pai setup' to configure storage backend, indexing, and daemon.",
"daemon": "Run 'pai daemon install' to start the background indexing service."
},
"notes": "Gemini CLI supports MCP servers. PAI's 9 MCP tools work fully. Hooks and skills are Claude Code-specific features that require Claude Code."
}
What ships with it
19 files beside gemini-extension.json in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- examples/workers.yaml 2.3 KB
- scripts/bench-embed.mjs 1.6 KB runs code
- scripts/bench-sort.mjs 927 B runs code
- scripts/build-docs.mjs 11 KB runs code
- scripts/build-hooks.mjs 11 KB runs code
- scripts/build-skill-stubs.mjs 6.3 KB runs code
- scripts/embed-ollama-job.mjs 11 KB runs code
- scripts/lib/sync-guard.mjs 5.3 KB runs code
- scripts/lib/sync-guard.test.mjs 4.6 KB runs code
- scripts/probe-cache-keepalive.ts 2.7 KB runs code
- scripts/proof-daemon-preload.mjs 2.0 KB runs code
- scripts/prune-dist.mjs 4.5 KB runs code
- templates/agent-prefs.example.md 10 KB
- templates/ai-steering-rules.template.md 1.7 KB
- templates/claude-md.template.md 23 KB
- templates/pai-project.template.md 373 B
- templates/pai-skill.template.md 16 KB
- templates/skills/user/.gitkeep 0 B
- templates/voices.example.json 8.6 KB
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 16d ago Changed e03f42174e07
- 1mo ago First seen · 27 lines scan A 55eb7cc5910f
@tekmidian/pai is a plugin published in the GitHub repository mnott/PAI (52 stars, last pushed 12d ago), licensed MIT. Its token cost is not measured: this kind of file is read by the harness, not the model. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-06.
Other plugins, from other repositories
open-knowledge-vault
Create and maintain portable Open Knowledge Format vaults with optional Obsidian and Dataview views.
knowledge-and-memory
Research methodology, corpus search, persistent memory, iterative workflows, and project knowledge management.
tree-ring-memory
Local-first memory lifecycle, project bootstrap, and receipt-backed harness guidance for Claude Code using Tree Ring Memory v0.15+.
claude-mega-brain
OKF-powered knowledge context for Claude Code — injects your project knowledge base at every session.
vault-template
Personal Knowledge Management system for Obsidian with AI-powered workflows, goal tracking, and productivity coaching.
bestpractice
Enforcement, memory and parallel-session coordination for solo founders running several Claude Code sessions on one repository.