Senior engineer
Build, reason, ship.
Role kits
Product designer
Turn intent into a polished interface.
SEO & marketer
Find demand, write for humans, distribute.
QA tester
Check the product, not just the code.
Product / project manager
Keep decisions, delivery and team context connected.
Docs team
Make technical work clear and reusable.
Social manager
Plan, create and publish with a consistent voice.
Motion designer
Create video, motion and post-production faster.
Researcher
Stay close to the signal and recent evidence.
The whole agency
Multi-role playbooks for an AI-native team.
Catalog
Tool catalogue
Orchestration & multi-agent runners
8○ queued
llm-council
A council of several LLMs that answer hard questions together.
○ queued
symphony
Turns project tasks into isolated autonomous runs so teams manage work instead of babysitting coding agents.
◐ testing
hermes-agent
Self-improving agent from Nous Research with a built-in learning loop; desktop build, MIT.
○ queued
paperclip
Open app for managing working agents.
○ queued
nasiko
Developer control plane: a dashboard for your own AI agents.
○ queued
agency-agents
Ready-made set of digital-agency agents (frontend, community, review), each with its own role and process.
○ queued
openclaude
Open terminal agent that runs anywhere and uses any LLM.
◐ testing
repo-task-proof-loop
Spec-driven skill that spawns subagents for repository tasks.
Model access
2● in use
OmniRoute
Free MIT gateway: one endpoint, 290+ providers (90+ free), 500+ models; quota fallback, context compression, MCP/A2A.
○ queued
gpt4free
Collection of free/bypass providers for language-model access. Legal status and stability of such proxies are unverified.
Context, memory, index, graph
8● in use
context7
Platform that feeds up-to-date library documentation to LLMs and AI editors.
● in use
claude-mem
Cross-session memory: records agent actions, compresses them with AI, and injects relevant context into later sessions.
○ queued
GitNexus
Client-side (in-browser, no server) knowledge-graph builder for a repository with a built-in Graph-RAG agent.
◐ testing
Understand-Anything
Turns a codebase into an interactive knowledge graph you can search and question.
○ queued
codebase-memory-mcp
MCP server that indexes a codebase into a persistent knowledge graph: 158 languages, sub-millisecond queries.
◐ testing
beads
A memory upgrade for your coding agent.
○ queued
hindsight
Agent memory that learns from experience.
○ queued
chroma-mcp
MCP server giving an agent access to the Chroma vector database.
Token optimization
1● in use
headroom
Compresses tool output, logs, files and RAG chunks before they reach the LLM (claimed ~20% for coding agents, 60–95% on JSON). Library, proxy, MCP server.
RAG & knowledge bases
1○ queued
maxkb
Enterprise-grade open-source platform for building agents on top of a knowledge base.
Search
1◐ testing
searxng
Free metasearch engine aggregating results from many engines without tracking users.
Security
3◐ testing
strix
Open-source AI pentest tool: finds application vulnerabilities and helps fix them.
● in use
skill-scanner
Security scanner for agent skills: check what is inside before installing.
○ queued
prowler
Open cloud-security platform: automates security and compliance checks across cloud environments.
Claude Code: kits & guides
6○ queued
agent-skills
Production-grade engineering skills for AI coding agents, covering the lifecycle from specification and planning through testing, review and release.
○ queued
gstack
Opinionated Claude Code setup with 23 tools spanning CEO, designer, engineering manager, release manager, documentation and QA roles.
◐ testing
ECC
Agent-harness optimization system: skills, instincts, memory, security, research-first approach. Formerly everything-claude-code.
◐ testing
awesome-claude-code-toolkit
Large collection: 135 agents, 35 skills, 42 commands, 176+ plugins, hooks, rules, MCP configs.
○ queued
SuperClaude_Framework
Configuration framework: specialized commands, cognitive personas, development methodologies.
○ queued
Claude-Code-Everything-You-Need-to-Know
Practical guide: setup, prompt engineering, slash commands, skills, hooks, subagents, MCP.
Thinking & output skills
3● in use
adhd
Tree-of-thought with pruning on the Agent SDK: parallel divergence across cognitive frames, then scoring and deepening the survivors.
○ queued
i-have-adhd
Skill that stops the agent burying the answer: ADHD-friendly output format.
◐ testing
no-ai-slop
Strips 20+ common AI-filler patterns out of text.
Applied agents & content
6◐ testing
open-design
Local desktop analogue of Claude Design: a coding agent as a design engine (prototypes, landings, dashboards, slides, video) with HTML/PDF/PPTX/MP4 export.
○ queued
awesome-design-md
Collection of ready-to-use DESIGN.md files analyzing popular websites so AI agents can generate consistent UI.
○ queued
diagram-design
Skill for editorial diagrams (29 types per the description, 27 per the README), outputting HTML + SVG instead of Mermaid. Picks up brand style from a site and redraws draw.io files.
○ queued
remotion-video-director
Interactive Claude Code skill for making Remotion videos through an expert discussion.
○ queued
instagram-cli
Agent (and MCP) for working with Instagram from the CLI.
○ queued
open-seo
Open alternative to Semrush and Ahrefs.
Execution infrastructure
1○ queued
OpenSandbox
Secure, fast and extensible sandbox runtime for AI agents.
Research & models
3○ queued
autoresearch
AI agents autonomously running research on training nanochat on a single GPU.
○ queued
MiroFish
Simple general-purpose swarm-intelligence engine for forecasting.
◐ testing
personaplex
Real-time full-duplex speech-to-speech model with persona control via a role prompt and voice audio conditioning (Moshi architecture).