A complete AI agency at your fingertips - From frontend wizards to Reddit community ninjas, from whimsy injectors to reality checkers. Each agent is a specialized expert with personality, processes, and proven deliverables.
130k
Stars
21.3k
Forks
95
Open issues
87
Contributors
AI Analysis
Agency Agents is a curated collection of specialized AI agent personality definitions (primarily Markdown files) designed to be loaded into AI coding tools like Claude Code, Cursor, Codex, and Gemini CLI to give them focused roles such as frontend developer, community manager, or content editor. It serves developers and AI power users who want pre-built, opinionated agent personas rather than writing their own system prompts from scratch. It is not for users seeking a general-purpose AI frame...
Inferred from signals mentioned in the README (tests, CI, type safety) — not a review of the actual code.
AI's overall editorial judgment — not an average of the bars above, can weigh other factors too.
Curated AI agent persona library with 100k+ stars and a native desktop installer
agency-agents is a collection of pre-written AI agent persona definitions (markdown files) covering engineering, marketing, security, and other domains, designed to be loaded into AI coding tools like Claude Code, Cursor, Codex, and Gemini CLI. It targets developers and teams who want to configure AI assistants with role-specific behavior without writing custom system prompts from scratch. A companion native desktop app simplifies browsing and installing agents across tools. The project's momentum is substantial, with 118k+ stars and a Chinese-language fork (agency-agents-zh) indicating international reach.
Created in October 2025, apparently originating from a Reddit thread. Grew rapidly through 2025-2026 alongside the explosion of AI coding assistant tools like Claude Code and Cursor that support custom agent/persona configurations.
Growth appears driven by the simultaneous rise of agentic AI coding tools that accept custom persona definitions, making this a zero-infrastructure solution to a common pain point. The native app launch and Homebrew cask likely accelerated adoption by removing the friction of cloning and scripting. 3,486 stars in the last 7 days suggests a recent viral moment or featured placement.
The existence of a Homebrew cask, a native cross-platform desktop app, and a Chinese-language fork with 15,973 stars suggests real-world usage beyond casual exploration. The README references specific upstream bugs in supported tools, implying the maintainer has production-grade feedback. However, direct evidence of enterprise or team-scale deployment is not documented in the README.
Appears to be a flat collection of markdown files organized into division directories (engineering, security, etc.), paired with Shell scripts for installation and conversion. Likely no complex runtime logic — the install.sh script copies .md files into tool-specific configuration directories. A separate companion app repo (agency-agents-app) handles the GUI installer.
not documented in README
Last push was 2026-06-29, one day before the evaluation date, indicating very active maintenance. The README documents tool-specific bugs (e.g., OpenCode's 119-agent limit with a linked upstream issue), suggesting the maintainer is tracking real-world integration problems closely.
ADOPT IF: you use Claude Code, Cursor, Codex, or other supported AI tools and want well-structured role-specific system prompts without writing them yourself — the native app makes this nearly zero-effort. AVOID IF: you need programmatic, dynamic, or runtime-configurable agent behavior, as this is a static file collection and not a framework. MONITOR IF: you are evaluating whether the agent quality and persona depth justify the hype, since star counts reflect distribution success but not necessarily prompt effectiveness in practice.
Independent dimensions
Mainstream potential
7/10
Technical importance
4/10
Adoption evidence
5/10
- The core artifact is markdown text files — the actual value depends entirely on prompt quality, which cannot be verified from metadata alone. Stars measure distribution, not effectiveness.
- Heavy dependency on third-party tool compatibility: if Claude Code, Cursor, or Codex change how they load agent definitions, the entire library could break or require mass updates.
- The niche may commoditize quickly — as AI tool vendors ship their own curated agent libraries, the differentiation of a third-party collection may erode.
- No documented quality assurance or test coverage for agent outputs means users have no benchmark for whether a given agent persona actually performs better than a generic prompt.
- Rapid star growth combined with a very young repo (created Oct 2025) may partly reflect trend-following rather than validated production utility — adoption depth is not independently verifiable.
Likely to remain a popular reference collection as long as file-based agent configuration remains the dominant pattern in AI coding tools. May face pressure from first-party agent marketplaces from Anthropic, Google, or OpenAI within 12-18 months.
Newsletter
Get analyses like this every Monday
Free weekly digest of the most interesting open-source discoveries.
Languages
Information
- Language
- Shell
- License
- MIT
- Last updated
- 16h ago
- Created
- 9mo ago
- Analyzed with
- anthropic/claude-sonnet-4-6
Stars over time
Contributors over time
Top 100 contributors only — repos with more will plateau at 100.
Open issues
Top contributors
Recent releases
No releases published yet.
Similar repos
kyrolabs/awesome-agents
Awesome Agents is a curated directory of open-source tools, frameworks, and...
contains-studio/agents
Contains Studio AI Agents is a collection of 40+ specialized AI agents...
jnMetaCode/agency-agents-zh
A Chinese-localized collection of 266 pre-built AI agent personas designed for...
777genius/agent-teams-ai
Agent Teams AI is a desktop application that orchestrates multiple AI agents to...
| Repository | Stars | Week Δ | Language | Score | Updated |
|---|---|---|---|---|---|
|
msitarzewski/agency-agents
this
|
130k | +3.5k | Shell | 7/10 | 16h ago |
|
kyrolabs/awesome-agents
|
2.6k | — | — | 7/10 | 14h ago |
|
contains-studio/agents
|
12.4k | — | — | 6/10 | 12mo ago |
|
jnMetaCode/agency-agents-zh
|
17k | — | Shell | 7/10 | 3d ago |
|
777genius/agent-teams-ai
|
1.6k | — | TypeScript | 7/10 | 6h ago |
 addyosmani/agent-skills
|
76.5k | — | JavaScript | 8/10 | 9h ago |
67,954 stars and Shell-based like this project, suggesting similar positioning. Likely the closest competitor in the same niche. Without README access it's unclear whether scope or agent quality differs materially.
A direct Chinese-language fork with 15,973 stars, indicating significant adoption in Chinese-speaking developer communities rather than a competing approach.
12,395 stars with no listed language, likely a similar collection. Smaller reach suggests agency-agents has stronger distribution or broader tool support.
TypeScript-based at 5,909 stars. May target a more programmatic or runtime-integrated approach rather than static markdown persona files.
Python-based at 2,122 stars, likely targeting a different integration pattern (scripted/API-driven) rather than file-drop persona configuration.
