Built for AI developers, technical founders, and engineering leads. RepoRadar separates discovery, editorial review, installation, bounded representative workflows, later-dated re-testing, and retained operator use; scores never promote a product into a hands-on stage.
Every record is reviewed or backed by an eligible retained run, failed-only attempts and discovery-only records are excluded, and the shortlist is balanced across technical categories. The published score floor is 7.0; explicit editorial or hands-on evidence may override it.
46 canonical entities currently have a retained passing bounded representative workflow run inside the credential-stripped, disposable, no-paid-service harness. This does not establish production use. catalog score orders only the eligible pool; popularity and entity ID are deterministic tie-breakers. The JSON contract lists exact run IDs, evidence URLs, category coverage, catalog-rank overlap, and evidence-backed high-ranked exclusions.
Commercial facts are dated and first-party sourced. Open-source records retain their direct repository source.
AG2 is an active open-source AgentOS with protocol-level tools, multi-agent orchestration, and explicit A2A interoperability design.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:retested recently · target category:agent framework · quality:score at or above 7.0
FastMCP is Prefect's Apache-2.0 Python framework for building MCP servers and clients with a high-level API, transport support, server composition, auth hooks, and deployment paths that reduce the boilerplate around exposing tools to agent clients.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:retested recently · target category:mcp infrastructure · quality:score at or above 7.0
OpenAI's Codex is an official, open-source terminal coding agent that turns prompts into concrete code operations (open/edit/run/test), integrates approval-aware execution, and supports multiple AI providers via OpenAI-compatible APIs.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · target category:coding agent · quality:score at or above 7.0
Gemini CLI is Google's Apache-2.0 terminal agent that gives developers direct Gemini access with a free tier, 1M-token context, built-in shell and file tools, web fetching, and MCP extensibility.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · target category:coding agent · quality:score at or above 7.0
Kilo Code is an MIT-licensed open-source coding agent that runs in VS Code, JetBrains, and the CLI, lets users pick from 500+ models with provider pricing, and supports switching models mid-task.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · target category:coding agent · quality:score at or above 7.0
serena is an MIT-licensed semantic coding MCP toolkit that gives Claude Code, Codex, and other agent clients symbol-aware retrieval, structured edits, and project memory, so coding assistants can work more like an IDE than a grep loop.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · target category:coding agent · quality:score at or above 7.0
Fission-AI/OpenSpec is an MIT spec-driven development framework and CLI for AI coding assistants that turns vague requests into reviewable proposals, plans, task artifacts, and archive history, with workflows aimed at real brownfield codebases instead of toy demos.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · target category:coding agent · quality:score at or above 7.0
kucherenko/jscpd is an MIT-licensed copy/paste detector for source code that supports 224+ formats, ships a token-efficient reporter, and exposes an MCP server so AI coding agents can call it as a real tool to flag duplicate code in pull requests and large refactors.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · target category:coding agent · quality:score at or above 7.0
AgentScope is a production-oriented Python framework for multi-agent systems with explicit tooling, state handling, and MCP-compatible execution paths.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · target category:agent framework · quality:score at or above 7.0
OpenAI Agents SDK for Python is an MIT-licensed, provider-agnostic framework for building multi-agent and voice-agent workflows with tools, MCP support, handoffs, guardrails, tracing, sessions, sandbox agents, and human-in-the-loop controls, and the current v0.17.6 release adds pre-approval tool input guardrails and SDK-only custom data for tool outputs.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · target category:agent framework · quality:score at or above 7.0
OpenAI Agents SDK for JavaScript/TypeScript is an MIT-licensed, provider-agnostic framework for building multi-agent and voice-agent workflows with tools, MCP support, handoffs, guardrails, tracing, sessions, sandbox agents, and human-in-the-loop controls.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · target category:agent framework · quality:score at or above 7.0
Flue is an Apache-2.0 TypeScript agent harness that packages sessions, tools, skills, subagents, durable execution, filesystem access, and sandboxed runtime controls into one framework for autonomous AI workflows.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · target category:agent framework · quality:score at or above 7.0
Agno is a Python framework for building, running, and managing agent platforms with clear abstractions for tools, memory, routing, and observability.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · target category:agent framework · quality:score at or above 7.0
Semantic Kernel 1.43.1 adds Azure AI/OpenAI Assistant function-choice behavior support and hardens schema handling for OpenAPI plugins, improving compatibility for production-grade agent integrations.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · target category:agent framework · quality:score at or above 7.0
BerriAI/litellm is a production-ready proxy and toolkit that normalizes calls across 100+ LLM providers while adding observability and guardrails.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · target category:inference runtime · quality:score at or above 7.0
MIT Microsoft cross-platform ML inference and training accelerator with ONNX standard support + hardware accelerators across CPU / GPU / NPU; 21,042* at verify time, last commit 2026-07-08; install `pip install onnxruntime`
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · target category:inference runtime · quality:score at or above 7.0
Apache-2.0 the open-source search infrastructure for AI -- embedding database / search engine for LLM applications; dense + sparse + full-text + multi-vector search in one API; SQL-style metadata filtering, late-interaction / SPLADE / ColBERT rerankers, native LangChain / LlamaIndex / OpenAI / Claude / MCP integration; ships the official chroma-mcp server companion and Chroma Cloud / Enterprise
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · target category:rag knowledge · quality:score at or above 7.0
zvec is an Apache-2.0 lightweight, in-process vector database from Alibaba Group that ships production-grade low-latency similarity search with no server to run. The v0.5.0 release (June 12, 2026) adds native full-text search, hybrid vector+text queries via a single MultiQuery, a DiskANN on-disk index for large datasets, plus first-party Go and Rust SDKs alongside the Python and Node packages.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · target category:rag knowledge · quality:score at or above 7.0
neuml/txtai is an Apache-2.0-licensed, open-source all-in-one AI framework for semantic search, LLM orchestration, embeddings, and language-model workflows that ships as a Python library, an API server, and a vector+graph+relational backend stack, so AI application teams, semantic-search engineers, and AI agent builders can run a single OSS toolkit that covers retrieval, RAG, and LLM
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · target category:rag knowledge · quality:score at or above 7.0
Sibyl-Labs/Sibyl-Memory is the MIT local-first agent-memory family that ships as **four PyPI packages** (`sibyl-memory-client`, `sibyl-memory-cli`, `sibyl-memory-hermes`, `sibyl-memory-mcp`) and runs entirely on a single SQLite file with FTS5 search — **no vector store, no embeddings, no retrieval model, no external infrastructure**; the architecture is a **five-tier hierarchical schema** —
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · target category:rag knowledge · quality:score at or above 7.0
microsoft/presidio is an MIT-licensed, context-aware PII de-identification SDK from Microsoft that detects and anonymizes sensitive entities in text and images and ships pluggable recognizers so AI builders can scrub PII before sending data to LLMs or storing it in logs.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · target category:security · quality:score at or above 7.0
e2a is an authenticated email gateway for AI agents that verifies inbound senders with SPF and DKIM, signs deliveries with HMAC headers, supports webhook or WebSocket delivery, and can hold outbound mail behind a human approval gate.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · target category:business workflow · quality:score at or above 7.0
anomalyco/opencode is an MIT-licensed open-source coding agent that ships as a local CLI with a hosted UI, broad language docs, and a real install path instead of another prompt pack or thin wrapper around an existing editor assistant.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
Apache-2.0 canonical Agent Client Protocol (ACP) -- open standard (current stable protocol version 1) for connecting any code editor to any coding agent via JSON-RPC; ships first-party SDKs in 5 languages
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
awslabs/mcp is a maintained catalog of production AWS MCP servers with documented setup paths for retrieving and acting on AWS context safely from agent clients.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
Kreuzberg is a Rust-core document intelligence framework that extracts text, metadata, tables, OCR results, and code structure from 96 file formats and 306 programming languages, with bindings for major runtimes plus CLI, REST, and MCP surfaces.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
Uni-CLI is a computer-use control plane for agents that routes intent into governed actions across browsers, websites, desktop apps, local tools, files, operating systems, and MCP servers, then returns evidence and repair loops instead of a bare success or failure.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
Qwen Code is an Apache-2.0 open-source terminal coding agent with Auto-Memory, Auto-Skills, subagents, MCP support, IDE integrations, daemon mode, and support for hosted or local model backends instead of a single locked-in provider path.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
LlamaIndex is an MIT-licensed framework and document-agent platform for turning messy files into searchable, extractable, and agent-ready knowledge, with OSS components plus managed Parse and Extract services.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
MiMo-Code is Xiaomi's MIT-licensed terminal coding agent that pairs low-friction setup with multi-agent modes, checkpointed memory, provider-agnostic model support, and a compose workflow for spec-driven development.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
CodeGraph is a MIT-licensed local code knowledge graph for Claude Code, Codex, Cursor, Gemini, OpenCode, AntiGravity, Kiro, and Hermes Agent that pre-indexes repositories, auto-syncs on changes, and aims to cut token use and tool calls versus raw grep-and-read loops.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
Apache-2.0 (raw master/LICENSE.md opens with the standard Apache-2.0 body on 2026-07-08) the LLM evaluation framework: 30+ research-backed metrics (G-Eval, hallucination, answer relevancy, bias, toxicity, summarization, JSON correctness), unit-test-style `assert_test()` integration with pytest, red-teaming suite for safety/privacy/bias/competitor-mention, side-by-side A/B regression comparison
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
earendil-works/pi is an MIT agent toolkit that bundles a unified multi-provider LLM API, agent runtime, interactive coding-agent CLI, terminal UI library, and documented container or OpenShell isolation paths, giving builders one open stack for running serious terminal-first agent workflows instead of stitching together separate wrappers and SDKs.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
PPT Master is an MIT-licensed workflow that turns documents into natively editable PowerPoint decks with real shapes, animations, speaker-note narration, and optional template-following instead of flattened slide images.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
Langflow is an MIT visual and API framework for building AI workflows, agents, and MCP-aware applications. It gives teams a drag-and-drop editor, reusable components, Playground testing, custom Python components, flow serving through an API, Docker and deployment docs, and a PyPI package that currently tracks the 1.10.x line.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
yamadashy/repomix is a widely used MIT-licensed CLI and web tool that packs an entire codebase into one AI-friendly bundle, making it much easier to hand real repository context to Claude, ChatGPT, Gemini, DeepSeek, or other coding models without improvising a fragile manual export workflow.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
Apache-2.0 open-source AI observability + evaluation + optimization platform from Comet ML with full prototype-to-production loop (tracing LLM calls / tool calls / agent steps / RAG retrieval + rerank + generation; LLM-as-judge + heuristic + custom metrics + dataset management + experiment tracking; Opik Optimizer + Opik Agent Optimizer for automatic prompt + tool optimization); 8+ LLM provider
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
promptfoo is an MIT-licensed CLI and library for evaluating prompts, agents, and RAG systems, with declarative test configs, provider comparisons, red-team checks, and CI-friendly workflows for catching quality and safety regressions before release.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
pydantic/pydantic-ai is an MIT-licensed Python agent framework from the Pydantic team that gives developers a type-safe, model-agnostic, FastAPI-style way to build production LLM agents with first-class tool calling, structured outputs, and dependency injection.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
Haystack increments toward production-grade retrieval and agent workflows with pragmatic API parameter improvements and ongoing pipeline maintenance.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
vercel-labs/skills is an MIT CLI and runtime for the open agent skills ecosystem that lets one `npx skills add owner/repo` command install, update, and resolve skill packages across Claude Code, Codex, Cursor, OpenCode, Antigravity, and 68 more coding agents from the same GitHub shorthand, instead of forcing each agent's users to learn a different install path.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
LangChain is a Python/JavaScript agent-engineering framework for composing model calls, tools, retrieval, structured outputs, multi-agent flows, and provider integrations.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
DaoyuanLi2816/can-i-finetune-this is the MIT-licensed `canifinetune` PyPI package (https://pypi.org/project/canifinetune/) that answers, before the user spends 14 GB of disk and an OOM on step 1, the five questions every consumer-GPU fine-tuner has: (1) Can I fine-tune this model? (2) About how much VRAM will it use? (3) What batch size / sequence length / LoRA rank / quantization should I use?
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
data-context-hq/datacontext is an Apache-2.0 Python package for runtime attribution and context observability around data access by AI agents and applications, with installable extras for SQLAlchemy, Postgres, BigQuery, Dagster, Snowflake, and dbt.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
Ikalus1988/MisakaNet is an Apache-2.0 zero-dependency reference implementation of the Swarm Knowledge Protocol (SKP) — a shared experience substrate for AI agents in which one agent stalls on a failure, documents the workaround, and all agents skip that same failure path; the core engine (`misakanet-core` on PyPI) is pure Python stdlib (BM25 + tokenize + RRF fusion) with zero dependencies, the
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
LangGraph is a library (under the LangChain organization) for building stateful, multi-actor LLM applications as directed graphs.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:bounded workflow tested · technical fit override:hands on or commercial · quality:score at or above 7.0
Terminal coding agent that reads repositories, edits files, runs tools, and supports MCP and enterprise model gateways.
Pricing: Pro monthly: $20 month; Max 5x: $100 month; Max 20x: $200 month; Team annual: $25 user/month; Team monthly: $30 user/month
Published because: verification:installed and tested · target category:coding agent · quality:eligible hands on run
AI development suite spanning IDE completion, chat, CLI, code review, cloud agents, and GitHub-native workflows.
Pricing: Free: $0 month; Pro: $10 user/month; Pro Plus: $39 user/month; Max: $100 user/month
Published because: verification:installed and tested · target category:coding agent · quality:eligible hands on run
Coding agent available through CLI, IDE, desktop app, web/cloud tasks, code review, SDK, and automation workflows.
Pricing: Eligible ChatGPT plans: plan-dependent inclusion; Flexible Codex credits: token-based credits
Published because: verification:installed and tested · target category:coding agent · quality:eligible hands on run
Ollama 0.30.10 makes Command A / North family models run on Apple Silicon through MLX and updates the embedded llama.cpp build, improving local model coverage on common Mac hardware.
Pricing: Open-source terms; operating costs depend on hosting and model providers.
Published because: verification:installed and tested · target category:inference runtime · quality:score at or above 7.0