APIs, SDKs, repositories, eval tooling, frameworks, and builder platforms.
A practical directory of coding agents, agentic IDEs, app builders, frameworks, evaluation systems, and inference tools. Links point to official project or company pages; Pagish summaries are original and kept concise to avoid republishing vendor copy.
Tools that inspect repos, edit files, run commands, create diffs, and help developers move from prompt to reviewed code.
multi-agent software work across local apps, CLI, IDE, cloud tasks, reviews, and long-running implementation
Strong fit when teams want supervised coding agents with logs, tests, diffs, and ChatGPT account integration.terminal-native Claude coding workflows, repo edits, debugging, tests, and model-context-driven development
Good option for developers who already use Claude and want a command-line coding agent with local project context.AI-first code editing, codebase chat, autocomplete, agent edits, and VS Code-like workflows
Useful when developers want an AI-native editor rather than a separate terminal agent.IDE completions, chat, CLI help, code review, GitHub-native agent workflows, and PR-oriented software work
Strong default for teams already centered on GitHub, especially with Copilot agent and review workflows.agentic IDE workflows, Cascade, multi-agent command center, and model-flexible code editing
Worth tracking for teams comparing agentic IDEs and cloud/local agent coordination.autonomous cloud software engineering, backlog tasks, migrations, PRs, tests, and multi-repo projects
Best evaluated on scoped engineering tasks where the team can review plans, logs, and pull requests.fast editor-native agent workflows, external agents, MCP, and real-time review of AI edits
Relevant for developers who care about editor speed and want to bring agents such as Codex, Claude, or OpenCode into Zed.agentic coding in VS Code-compatible editors and the terminal, with team collaboration around threads
Useful for teams pairing AI coding with Sourcegraph code intelligence and MCP integrations.Open-source or source-available agents developers can run, inspect, extend, or pair with multiple model providers.
VS Code, JetBrains, CLI, SDK, MCP, multi-agent, and human-approved agent edits
One of the most active open-source coding-agent projects, with broad model-provider support and extension points.terminal, desktop, and IDE coding-agent workflows with many model providers and LSP support
Good fit for developers who want an open-source agent with local workflows and provider flexibility.terminal pair programming, git-native edits, test/lint loops, and multi-language codebases
Durable option for developers who want transparent terminal workflows and git commits from agent changes.VS Code and JetBrains chat, edit, autocomplete, agent mode, and custom AI code agents
Continue joined Cursor, but its open-source codebase remains important for teams studying extensible assistant patterns.local desktop/CLI agents, MCP extensions, workflow automation, code, data, and general technical tasks
Important because it is moving under the Agentic AI Foundation and emphasizes local extensibility.Prompt-to-app tools for prototypes, web apps, hosted IDEs, and product teams that need working software quickly.
browser IDE app building, hosted environments, deployments, and non-local development loops
Useful for quick full-stack prototypes and teams that value an integrated hosted dev environment.prompt-to-app web development, prototypes, front-end apps, and one-click hosted previews
Good for product exploration and early-stage apps where speed matters more than deep repo control.prompt-driven full-stack web apps, product prototypes, UI iteration, and Supabase-style app workflows
Relevant for non-specialist builders and small teams shipping web app MVPs quickly.UI generation, React components, design-to-code, and Vercel-oriented front-end prototyping
Strong signal for interface prototyping, especially for teams already using React and Vercel.Libraries and SDKs for building AI agents, RAG apps, tool use, workflows, and production AI applications.
agent apps, tools, handoffs, guardrails, tracing, and OpenAI-model workflows
Good starting point when building structured agents rather than only calling a chat API.agent orchestration, LangGraph workflows, retrieval, observability, and production AI app patterns
Still a major ecosystem signal for agent frameworks and production workflow patterns.RAG, data agents, document pipelines, indexes, retrieval, and context engineering
Strong fit for teams whose bottleneck is connecting private data to models.streaming AI UI, chat apps, tool calling, model providers, React/Next.js AI apps, and serverless deployments
Practical default for web developers building AI interfaces in TypeScript.agent development, multi-agent systems, tools, evaluation, deployment, and Gemini/Google Cloud workflows
Relevant for teams building enterprise agents around Gemini and Google Cloud.Tools for evaluating, tracing, monitoring, debugging, and improving AI applications after the first demo works.
tracing, evaluation, prompt testing, datasets, and debugging LLM apps and agents
Important when teams need repeatable evals and trace visibility for production AI workflows.LLM tracing, evaluations, monitoring, dataset inspection, and model/application debugging
Strong option for ML teams that already use W&B and need AI app observability.AI evals, prompt iteration, datasets, experiments, observability, and human review workflows
Useful for teams moving from subjective prompt testing to tracked evaluation loops.LLM gateway observability, logging, cost tracking, caching, rate limits, and provider monitoring
Good for teams that want lightweight LLM telemetry across multiple providers.Developer tools for running models locally, serving models, routing providers, and deploying inference workloads.
local model serving, experimentation, embeddings, and developer-friendly model downloads
Useful when privacy, offline testing, or local model iteration matters.local model chat, OpenAI-compatible local server workflows, model testing, and desktop experimentation
Practical for developers testing local models before wiring them into apps.high-throughput LLM serving, paged attention, batching, prefix caching, and production inference
Important for teams serving open models at scale.model routing, OpenAI-compatible proxying, provider abstraction, budgets, logs, and fallbacks
Useful when teams need a stable interface across many model providers.hosted model APIs, image/video/audio models, demos, deployment, and model discovery
Helpful for developers who need quick API access to many open models without managing infrastructure.Frameworks, hosted inference, observability, vector retrieval, and eval harnesses are becoming the practical layer between model releases and production apps.
SDK adoption, repository velocity, model-provider support, and repeatable evaluation workflows.Stars alone are noisy. Pagish weighs release cadence, issues, forks, maintainers, examples, and whether practitioners are actually building with a project.
New releases, docs quality, community examples, and production integrations.Tool choices are being shaped by latency, batch APIs, caching, quantization, GPU availability, and the ability to swap models without rewriting apps.
Pricing, rate limits, model catalogs, structured outputs, and deployment portability.3Blue1Brown — YouTube is listed in the AI source profile. Reference watchlist URL only; not fetched by runtime unless a structured feed or official API connector is added.
source500 Global500 Global is listed in the Business source profile. Reference watchlist URL only; not fetched by runtime unless a structured feed or official API connector is added.
sourceAAAIAAAI is listed in the Technology source profile. Reference watchlist URL only; not fetched by runtime unless a structured feed or official API connector is added.
sourceABB RoboticsABB Robotics is listed in the Technology source profile. Reference watchlist URL only; not fetched by runtime unless a structured feed or official API connector is added.
sourceABC News (US)ABC News (US) is listed in the World source profile. Reference watchlist URL only; not fetched by runtime unless a structured feed or official API connector is added.
sourceABC News AustraliaABC News Australia is listed in the World source profile. Reference watchlist URL only; not fetched by runtime unless a structured feed or official API connector is added.
sourceAccelAccel is listed in the Business source profile. Reference watchlist URL only; not fetched by runtime unless a structured feed or official API connector is added.
sourceAdobe (ADBE)Adobe (ADBE) is listed in the Stocks source profile. Reference watchlist URL only; not fetched by runtime unless a structured feed or official API connector is added.
