🔭 On the radar
Up-and-coming candidates — surfaced by the weekly discovery scan or submitted by the community — that haven't cleared the curation bar or a vetting pass yet. Stars refresh weekly from the discovery queue; descriptions are the projects' own, unvetted. Entries graduate into the ranked list or drop off.
| Project | Stars | What it says it is | Via |
|---|---|---|---|
| agenticSeek | 27.6k | Fully Local Manus AI. No APIs, No $200 monthly bills. Enjoy an autonomous agent that thinks, browses the web, and code for the sole cost of electricity. | weekly discovery |
| openfang | 18.2k | Open-source Agent Operating System | weekly discovery |
| evolver | 8.7k | The GEP-powered self-evolving engine for AI agents. Auditable evolution with Genes, Capsules, and Events. \| evomap.ai | weekly discovery |
| ZCode | 8.4k | Z.ai's coding agent harness. Powerful, intelligent, extensible. | weekly discovery |
| reef | 7.8k | Infrastructure for continually self-improving agents: records live traffic and feedback, then ships a scored update to model weights or the harness. | community · PR #125 |
| openevolve | 7.7k | Open-source implementation of AlphaEvolve | weekly discovery |
| intentkit | 6.5k | IntentKit is an open-source, self-hosted cloud agent cluster that manages a collaborative team of AI agents for you. | weekly discovery |
| agentscope-java | 6.2k | Build distributed, production-grade, long-running agents. | weekly discovery |
| vibesdk | 5.4k | An open-source vibe coding platform that helps you build your own vibe-coding platform, built entirely on Cloudflare stack | weekly discovery |
| eko | 5k | Eko (Eko Keeps Operating) - Build Production-ready Agentic Workflow with Natural Language - eko.fellou.ai | weekly discovery |
| fast-agent | 4k | Code, Build and Evaluate agents - excellent Model and Skills/MCP/ACP/A2A Support | weekly discovery |
| antigravity-sdk-python | 3.9k | A Python library for building AI agents that leverage the full power of Google Antigravity. | weekly discovery |
| MIRIX | 3.5k | Mirix is a multi-agent personal assistant designed to track on-screen activities and answer user questions intelligently. By capturing real-time visual data… | weekly discovery |
| OrcaReplay | 178 | Record, replay, fork, and debug any agent run with any model; time travel for AI agents. | community · issue #109 |
| placecall | 105 | Skill and plugins that give Claude Code, Codex, Cursor, and Gemini CLI a hosted MCP server for phone calls to US businesses. | community · issue #143 |
| 5dive | 73 | Run a company of agents on a server you own: each agent is an official CLI (Claude Code, Codex, Pi) as a systemd service, sharing one task queue. | community · issue #129 |
| llm-wiki-cli | 53 | Proactive memory CLI for AI agents: recall, maintain, and evolve persistent, source-grounded knowledge across sessions. | community · issue #79 |
| red-team-blue-team-agent-fabric | 30 | AI agent security harness for adversarial testing: 600+ executable tests across MCP, A2A, x402/L402, and decision governance. | community · issue #72 |
| mcp-memory-server | 27 | MCP memory server backed by a hosted engine: recall that re-ranks on helped or misled feedback, plus shared rooms for multi-agent work. | community · issue #120 |
| model-citizen | 24 | Control plane for Claude Code and Codex: one checkout of rules, skills, and roles projected into each runtime, with hook guardrails and cost records. | community · issue #119 |
| Autobot | 23 | AGENTS.md operating contract, on-disk memory, and an independent completion validator for ChatGPT and Codex local projects on a Mac. | community · issue #126 |
| loopgate_harness | 21 | Run any CLI agent in a loop against specs: fresh context each iteration, the repo as durable memory, lint/types/tests/security enforced by git hooks so bad c… | community · PR #69 |
| GoodMemory | 17 | Local-first, auditable memory layer for AI apps and coding agents (Codex, Claude Code, MCP, HTTP). | community · issue #44 |
| teamcopilot | 14 | Deploy AI agents for your team to automate business workflows and coding. | community · PR #21 |
| nausicaa-harness | 12 | Treats an agent run as a dynamic topology of lanes instead of one linear loop; the model decides when to observe or fan out. | community · issue #108 |
| atlarix-releases | 9 | Atlarix: proprietary desktop agent workstation; hunk-by-hunk diff approval, per-OS sandboxed shell, lexical retrieval, any model. Release repo only. | community · issue #48 |
| dsh-mcp-lens | 8 | MCP tool search for large catalogs: 1,000 tools behind 2 schemas, exact-schema calls, allow/deny controls. | community · issue #81 |
| house-party-protocol | 5 | Evidence and approval contracts around Claude Code and Codex CLI: done is a recorded exit code, and a verdict from the builder's own lane is refused. | community · issue #124 |
| workkit | 4 | Claude Code plugin that runs work through GitHub Issues: an accepted spec gates each build, a blind verifier reviews, and hooks enforce the test gate. | community · issue #136 |
| laravel-ai-verify | 3 | Verification gate for coding agents on Laravel 12/13: runs only the checks a diff needs, fails weakened tests, and blocks Claude Code from stopping early. | community · issue #140 |
| yylo-benchmark | 2 | Longitudinal benchmark for coding-agent runs: sandboxed fresh-repo attempts, deterministic and blinded LLM judging, hash-bound evidence. | community · issue #117 |
| jet-browser | 2 | Portable WPE WebKit browser runtime for agents: one isolated process driven by ordered JSONL or MCP commands, with native input and screenshots. | community · issue #137 |
| Giovanni | 2 | Claude Code plugin for a daily digest loop: parallel source pulls by subagents, drift checks against a constitution in git, and lint limits on memory. | community · issue #139 |
| agentlink | 2 | Coding-agent harness inside VS Code: editor code intelligence as tools, reviewable diffs, an approval policy, sandboxed commands, resumable sessions. | community · issue #145 |
| wippaAGENT | 1 | Install-once OpenCode config: one orchestrator with edit and shell denied delegates to 12 specialist subagents behind audit and review gates. | community · PR #147 |
| mcp-guardian | — | MCP proxy for tool scoping and context compression: progressive discovery with 3 meta-tools. | community · issue #46 |
| minimal-harness | — | Dependency-free acceptance and handoff kernel for coding agents: runs configured checks and fails closed when Git or contract evidence goes stale. | community · issue #115 |