TokRepo - Where AI Experience Lives
Infrastructure for the AI era. Manage your skills, MCPs, prompts and scripts in one place — humans confirm intent, agents install or push through TokRepo tools.
Connect TokRepo through MCP or CLI so any capable agent can search, evaluate, plan, install, and push assets through TokRepo.
Agents call TokRepo tools directly
Search, detail, verify, install-plan, and push flows are structured contracts for MCP/CLI agents; the web page only hands off the contract.
tokrepo_search(query)tokrepo_detail(uuid)tokrepo_verify(uuid)tokrepo_install_plan(uuid)Standalone gems — hand-picked, refreshed every 4 hours
Author just shipped an update — back on the radar.
Worth a look in this round.
Surfacing this one this cycle.
Picked out of the pile this window.
One to check this time around.
Six GitHub-verified building blocks for terminal-first agent operations: session capture, an agent-aware multiplexer, a model/agent selector TUI, local memory, and two MCP directories.
- ▸Entire CLI — Record AI Sessions in Git
- ▸herdr — Terminal Multiplexer for Agents
- ▸models — TUI to Compare Models and Agents
Perplexity Sonar, Tavily, Exa, Firecrawl, Grok Live Search — the search APIs that ground AI agents in real-time web results with citations.
- ▸Perplexity Sonar API — Search-Grounded LLM in One Call
- ▸Perplexity Async Jobs — Long Research with Webhook Callbacks
- ▸Tavily Search — Search API Built for AI Agents
Six GitHub-verified building blocks to standardize agent skills, add persistent memory, and instrument runs end-to-end for measurable iteration.
- ▸MCP Memory Service — Persistent Agent Memory API
- ▸ToolHive — Run & Manage MCP Servers with thv
- ▸Skill Seekers — Convert Docs/Repos/PDFs to Skills
Promoted pack for practical MCP setups: a desktop client, Kubernetes + Argo CD ops, safer database querying, and skill/proxy layers you can copy into agents.
- ▸TUUI — Desktop MCP Client for Local Toolchains
- ▸kubernetes-mcp-server — Kubernetes Tools via MCP
- ▸Argo CD MCP — Manage Argo Apps via MCP
Promoted pack for safer tool-calling: prompt-injection defense, replayable automation, deterministic IAM drafts, and precise code indexing for agents.
- ▸Defender — Prompt Injection Defense for Tool Calls
- ▸Beelzebub — LLM Deception Runtime (MCP Honeypot)
- ▸flyto-core — Replayable Automation Engine (MCP)
Eight GitHub-verified tools to plan, run, and review coding agents: HUD visibility, plan/diff annotation, kanban runs, agentic IDE, code search MCP, and orchestration.
- ▸Claude Code Best Practice — Agentic Engineering Patterns
- ▸Claude Howto — Guided Claude Code Learning Path
- ▸Claude HUD — Statusline Dashboard for Claude Code
