AI IntelligencePUBLIC DISCOVERIES

IDEAS FOR AGENTIC DEVELOPMENT

Discover it.
Understand its value.

Saved public GitHub discoveries, explained in 50 words. Explore what each project does and how it could help you build, coordinate or evaluate agents.

Explore the feed ↓

EXPLORE THE FEED

Popular

50 on this page · Page 8

Projects ordered by their latest saved GitHub star count. Counts describe public interest, not quality. Check the dates on each card.

Loading your favourites…

  • Runtime patternsGitHub · Repository

    Nimbus

    AshishKumar4/Nimbus

    16 GitHub stars
    Checked 2026-10-10 UTC

    50-WORD AI EDITORIAL SUMMARY

    Nimbus provides POSIX-style sandbox infrastructure for agents running on Cloudflare. Potential use: Study execution loops and state handling before designing an agent runtime. This suggests an integration pattern; suitability depends on your tools and constraints. Before considering unattended use, check platform compatibility against your requirements in a small controlled test.

    Potential use

    Study execution loops and state handling before designing an agent runtime.

    Read full summary →View source ↗
    Publication date unavailableEditorial reviewed 2026-10-11

    About AI summaries

  • Task orchestrationGitHub · Repository

    detent

    digitaldrywood/detent

    16 GitHub stars
    Checked 2026-10-20 UTC

    50-WORD AI EDITORIAL SUMMARY

    Detent coordinates agent work using GitHub boards, isolated worktrees, validation gates and a serialized merge process. Potential use: Coordinate task handoffs with explicit responsibilities and review points. This suggests an integration pattern; suitability depends on your tools and constraints. Before considering unattended use, check permissions using a small controlled test.

    Potential use

    Coordinate task handoffs with explicit responsibilities and review points.

    Read full summary →View source ↗
    Publication date unavailableEditorial reviewed 2026-10-11

    About AI summaries

  • Reference & discoveryGitHub · Repository

    CoursIA

    jsboige/CoursIA

    16 GitHub stars
    Checked 2026-10-20 UTC

    50-WORD AI EDITORIAL SUMMARY

    CoursIA provides AI course materials and practical examples using C# and Python. Potential use: Compare documented approaches before choosing components for agent-assisted development. This suggests an integration pattern; suitability depends on your tools and constraints. Before considering unattended use, check platform compatibility against your requirements in a small controlled test.

    Potential use

    Compare documented approaches before choosing components for agent-assisted development.

    Read full summary →View source ↗
    Publication date unavailableEditorial reviewed 2026-10-11

    About AI summaries

  • Task orchestrationGitHub · Repository

    primer

    primerhq/primer

    16 GitHub stars
    Checked 2026-10-08 UTC

    50-WORD AI EDITORIAL SUMMARY

    Primer supplies a self-hosted control plane for agent fleets, workspaces, channels and workflow triggers. Potential use: Coordinate task handoffs with explicit responsibilities and review points. This suggests an integration pattern; suitability depends on your tools and constraints. Before integrating it, check permissions against your requirements in a small controlled test.

    Potential use

    Coordinate task handoffs with explicit responsibilities and review points.

    Read full summary →View source ↗
    Publication date unavailableEditorial reviewed 2026-10-11

    About AI summaries

  • Context & retrievalGitHub · Repository

    goldener

    goldener-data/goldener

    15 GitHub stars
    Checked 2026-10-08 UTC

    50-WORD AI EDITORIAL SUMMARY

    Goldener orchestrates AI pipeline data through reusable embeddings, sampling, annotation checks and distribution monitoring. Potential use: Supply relevant context to agents while preserving provenance and access boundaries. This suggests an integration pattern; suitability depends on your tools and constraints. Before considering unattended use, check permissions using a small controlled test.

    Potential use

    Supply relevant context to agents while preserving provenance and access boundaries.

    Read full summary →View source ↗
    Publication date unavailableEditorial reviewed 2026-10-11

    About AI summaries

  • Tool integrationGitHub · Repository

    franken_agent_detection

    Dicklesworthstone/franken_agent_detection

    14 GitHub stars
    Checked 2026-10-07 UTC

    50-WORD AI EDITORIAL SUMMARY

    Franken Agent Detection discovers installed coding-agent connectors through deterministic local filesystem checks. Potential use: Connect existing software capabilities to agent tools through explicit interfaces. This suggests an integration pattern; suitability depends on your tools and constraints. Before considering unattended use, check permissions against your requirements in a small controlled test.

    Potential use

    Connect existing software capabilities to agent tools through explicit interfaces.

    Read full summary →View source ↗
    Publication date unavailableEditorial reviewed 2026-10-11

    About AI summaries

  • Runtime patternsGitHub · Repository

    stuntman

    mhlaghari/stuntman

    14 GitHub stars
    Checked 2026-10-08 UTC

    50-WORD AI EDITORIAL SUMMARY

    Stuntman separates planning and review from execution across models in a coding-agent harness. Potential use: Study execution loops and state handling before designing an agent runtime. This suggests an integration pattern; suitability depends on your tools and constraints. Before adoption, check permissions against your requirements in a small controlled test.

    Potential use

    Study execution loops and state handling before designing an agent runtime.

    Read full summary →View source ↗
    Publication date unavailableEditorial reviewed 2026-10-11

    About AI summaries

  • Runtime patternsGitHub · Repository

    Mingbird-agent

    Mingbird/Mingbird-agent

    14 GitHub stars
    Checked 2026-10-07 UTC

    50-WORD AI EDITORIAL SUMMARY

    Mingbird Agent provides a Windows and Ollama harness aimed at smaller local models, with published benchmark material. Potential use: Study execution loops and state handling before designing an agent runtime. This suggests an integration pattern; suitability depends on your tools and constraints. Before integrating it, check permissions against your requirements.

    Potential use

    Study execution loops and state handling before designing an agent runtime.

    Read full summary →View source ↗
    Publication date unavailableEditorial reviewed 2026-10-11

    About AI summaries

  • Desktop executionGitHub · Repository

    protoAgent

    protoLabsAI/protoAgent

    14 GitHub stars
    Checked 2026-10-19 UTC

    50-WORD AI EDITORIAL SUMMARY

    ProtoAgent provides a local desktop agent workspace with coding-agent connections, plugins and agent communication. Potential use: Connect operator tasking to tools while preserving permission boundaries. This suggests an integration pattern; suitability depends on your tools and constraints. Before integrating it, check permissions against your requirements in a small controlled test.

    Potential use

    Connect operator tasking to tools while preserving permission boundaries.

    Read full summary →View source ↗
    Publication date unavailableEditorial reviewed 2026-10-11

    About AI summaries

  • Evaluation & controlsGitHub · Repository

    ElevenPowers

    SatyamSingh-Git/ElevenPowers

    14 GitHub stars
    Checked 2026-10-07 UTC

    50-WORD AI EDITORIAL SUMMARY

    ElevenPowers tracks evidence for correctness claims and invalidates completion when the underlying code changes. Potential use: Use measured evidence to distinguish completed work from unsupported agent claims. This suggests an integration pattern; suitability depends on your tools and constraints. Before considering unattended use, check permissions using a small controlled test.

    Potential use

    Use measured evidence to distinguish completed work from unsupported agent claims.

    Read full summary →View source ↗
    Publication date unavailableEditorial reviewed 2026-10-11

    About AI summaries

  • Evaluation & controlsGitHub · Repository

    agent-evaluation-framework

    Lin-chpin/agent-evaluation-framework

    13 GitHub stars
    Checked 2026-10-09 UTC

    50-WORD AI EDITORIAL SUMMARY

    Agent Evaluation Framework provides pluggable evaluation for agents and skills with auditable, reversible improvement processes. Potential use: Use measured evidence to distinguish completed work from unsupported agent claims. This suggests an integration pattern; suitability depends on your tools and constraints. Check permissions against your requirements in a small controlled test.

    Potential use

    Use measured evidence to distinguish completed work from unsupported agent claims.

    Read full summary →View source ↗
    Publication date unavailableEditorial reviewed 2026-10-11

    About AI summaries

  • Evaluation & controlsGitHub · Repository

    squidbrake

    batrapulkit/squidbrake

    12 GitHub stars
    Checked 2026-10-07 UTC

    50-WORD AI EDITORIAL SUMMARY

    Squidbrake checks tool calls against rules, requests approval for riskier actions and records an audit trail. Potential use: Use measured evidence to distinguish completed work from unsupported agent claims. This suggests an integration pattern; suitability depends on your tools and constraints. Before adoption, check permissions using a small controlled test.

    Potential use

    Use measured evidence to distinguish completed work from unsupported agent claims.

    Read full summary →View source ↗
    Publication date unavailableEditorial reviewed 2026-10-11

    About AI summaries

  • Evaluation & controlsGitHub · Repository

    humanish

    danielgwilson/humanish

    11 GitHub stars
    Checked 2026-10-08 UTC

    50-WORD AI EDITORIAL SUMMARY

    Humanish uses synthetic users to walk through software interfaces and exercise product interactions. Potential use: Use measured evidence to distinguish completed work from unsupported agent claims. This suggests an integration pattern; suitability depends on your tools and constraints. Before adoption, check permissions against your requirements in a small controlled test.

    Potential use

    Use measured evidence to distinguish completed work from unsupported agent claims.

    Read full summary →View source ↗
    Publication date unavailableEditorial reviewed 2026-10-11

    About AI summaries