Agent Systems
Reusable infrastructure for building and operating AI agents: runtimes, memory, context, tools, skills, coordination, and deployment. Reusable agent memory and general skill catalogs remain here even when coding-focused. Excludes complete end-user assistants, interaction-focused interfaces, coding agents and coding-specific workspaces, orchestration or client extensions, physical robot control, model execution, and security-primary tools.
Subfields
Agent Context
Tools that select, compact, transform, retrieve, or hand off the information supplied to an agent; it excludes long-term memory storage itself and model context-length limits at the hardware level.
5 projectsAgent Memory
Systems that persist, retrieve, organize, or learn from an agent's experiences, decisions, preferences, and prior sessions; it excludes ordinary document search and model context-window management alone.
6 projectsAgent Platforms
Broad self-hosted platforms for deploying, administering, governing, or packaging multiple agents and their tools; it excludes low-level agent loops, coding IDEs, and model-inference servers.
2 projectsAgent Runtimes
Reusable software whose primary product executes agent reasoning, tool-use, state, and approval loops. Complete personal assistants belong under personal-ai-applications even if they contain such loops. Among runtime leaves, recovery or replay as the main product takes precedence over embedding, then general harnesses. Excludes model execution and domain applications.
1 projectsAgent Skills
Reusable installable capability packs, instructions, procedural workflows, and catalogs extending agents. Excludes standalone applications, runtime frameworks, and security-primary skill inspection or verified-update tools, which belong under ai-security.
3 projectsMCP Agent Tools
Products primarily delivering MCP tool servers, transport gateways, proxies, or reusable agent tool bundles. Security-primary products belong under ai-security regardless of MCP packaging. Excludes domain applications with optional MCP adapters, ordinary service connectors, and instruction-only skill packages.
4 projectsMulti-Agent Systems
Reusable frameworks and protocols for agents to divide work, communicate, and coordinate. Communication-only products go to agent-communication; dynamic decentralized swarm coordination takes precedence over general multi-agent-frameworks. Excludes coding-specific orchestration, complete domain applications, and single-agent tool loops.
2 projects| Project | What it does | Stars | Language | License | Latest release |
|---|---|---|---|---|---|
| paperclipai/paperclip | An open-source platform for managing work agents with goals, budgets, permissions, and adapter-based orchestration. | ★ 103.9K | TypeScript | MIT | v2026.1012.0 |
| thedotmack/claude-mem | Persistent memory captures and compresses agent sessions, then injects relevant context into future sessions. | ★ 101.5K | TypeScript | Apache-2.0 | v14.0.1 |
| tt-a1i/archify | An agent skill that turns ideas, plans, or codebases into explorable interactive HTML architecture diagrams. | ★ 85.2K | JavaScript | MIT | v3.1.0 |
| rtk-ai/rtk | A CLI proxy that compresses, routes, and filters development-command output to reduce LLM token consumption in agent context. | ★ 83.8K | Rust | Apache-2.0 | v1.0.1 |
| headroomlabs-ai/headroom | Compresses tool outputs, logs, files, and RAG chunks before they enter an LLM context. | ★ 75.7K | Python | Apache-2.0 | v0.41.0 |
| hkuds/cli-anything | Provides installable CLI harnesses that let agents operate many software applications through structured interfaces. | ★ 54.4K | Python | Apache-2.0 | v0.4.0 |
| deusdata/codebase-memory-mcp | Indexes codebases into a persistent knowledge graph and provides code intelligence through MCP tools. | ★ 46.8K | C | MIT | v0.12.1 |
| getzep/graphiti | Persistent agent memory integrates changing facts and interactions with provenance. | ★ 32K | Python | Apache-2.0 | v0.31.1 |
| mksglu/context-mode | A context tool that sandboxes coding-agent output before inserting smaller representations into prompts. | ★ 30.6K | TypeScript | NOASSERTION | v1.0.169 |
| earthtojake/text-to-cad | Workflow skills that let agents create CAD models, check manufacturability, and produce engineering drawings. | ★ 21.4K | Python | MIT | v0.7.22 |
| kortix-ai/suna | Suna is a self-hosted AI platform packaging agents, skills, company memory, and connectors with isolated cloud workspaces and human approvals. | ★ 20.3K | TypeScript | NOASSERTION | v0.14.1 |
| tigerless-labs/agent-memory | A Markdown-based long-term agent memory runtime with local ranked retrieval and a sleep-time management layer. | ★ 4.2K | Python | MIT | — |
| oldcircle/geo-sleuth | A reusable agent skill geolocates photos using maps, elevation, satellite imagery, and Street View. | ★ 3.3K | Python | MIT | — |
| dzhng/jevgrep | Finds relevant files and source excerpts from natural-language descriptions for coding agents. | ★ 2.8K | TypeScript | MIT | v0.8.0 |
| duty1g/x64dbg-mcp-server | Provides x64dbg debugging and control operations through an HTTP-based MCP server. | ★ 2.5K | Zig | MIT | v1.6 |
| redhat-et/ripwire | Lets coding agents find relevant code, assess change impact, and select tests without reading the entire repository. | ★ 2.5K | C++ | Apache-2.0 | v0.6.7 |
| reason-machines/reason-network | A protocol engine for coding-agent sessions to exchange messages through direct P2P links or encrypted relays. | ★ 361 | Rust | Apache-2.0 | v0.10.5 |
| nousresearch/hermes-official-plugins | Official plugins adding Home Assistant, Spotify, and Claude SDK integrations to Hermes Agent. | ★ 303 | Python | MIT | — |
| graphicle-ai/graphicle | A local Rust tool that indexes code and stores sourced context from documents, decisions, and past sessions. | ★ 280 | Rust | Apache-2.0 | v0.1.2 |
| 100yenadmin/hermes-teammates | A Hermes plugin adding named subagent roles, context-carried follow-ups, live steering, and concurrent teammate execution. | ★ 79 | Python | MIT | v1.0.0 |
| doronp/gitmemory | Preserves an agent’s past conversations in Git and retrieves them by byte offset for long-term memory. | ★ 20 | Python | Apache-2.0 | v0.1.0 |
| nabzx/mnemosyne | Version control for agent memory with provenance, snapshots, branching, merging, blame, bisect, and time travel. | ★ 17 | Rust | Apache-2.0 | v0.0.8 |
| abird-ai/agentc | Agentc provides an embeddable AI agent in freestanding C23 under 1 MB, with C ABI and MCP support. | ★ 9 | C | MIT | v0.8.0 |