AI Coding

Perplexity open-source sandbox platform for AI agents and code-running applications. Create isolated Firecracker microVMs through an HTTP API, execute commands, stream output, and manage pause, resume, snapshots, and recovery on your own infrastructure. Apache-2.0 software; hosting and operations are your responsibility.

Task-specific AI API for classifying content, extracting structured fields, summarizing documents, and compressing agent context. Text input costs $0.05 per million tokens, supported image input costs $0.15 per million, and output is free; the current launch offer gives new accounts 50 million free tokens.

Prime Intellect's open-source, Rust-rewritten coding and research agent harness that uses recursive subagents, a persistent Python REPL, daemon-backed sessions, and a Continual Harness — with 14x faster input latency, 80%+ lower memory, and new Windows beta support.

Anthropic service that gives eligible open-source projects free, periodic AI vulnerability scans, emailing model-generated reports with reproducers and candidate patches to verified maintainers.

Vela 2.0 is a family of Apache-2.0 open routing models from the vLLM Semantic Router project that answers routing, safety, PII, and groundedness questions in one request and returns span-level decisions with probabilities.

Mem0 is an open-source AI memory layer that lets assistants and agents remember user preferences across sessions, with a pip/npm library, a self-hostable server, and a managed cloud platform.

Microsoft's open-source framework for defining MCP tool environments, running LLM agents with simulated users, and evaluating final backend state, side effects, and response requirements for reliable agent workflows.

Agent Skills is an MIT-licensed collection of production-grade workflows for AI coding agents, with lifecycle commands and reusable skills covering specification, planning, implementation, testing, review, and shipping.

Kolibri is Aleph Alpha's Apache-2.0 open-weight German-English mixture-of-experts reasoning model with 78B total and 3.46B active parameters, up to 1M-token context, tool calling, and self-hosted vLLM deployment.

Spark-X2.5 is an Apache-2.0 open model series with compact 4B and 1.7B variants, native 1M-token context, coding and agent capabilities, and local deployment across Ollama, LM Studio, vLLM, SGLang, llama.cpp, and MLX.

Production inference platform from Prime Intellect offering OpenAI-compatible access to hosted and gateway models, with serverless endpoints, reserved capacity, multi-datacenter failover, and infrastructure for long-running agent workloads.

Pricing engine and directory that compares 366 cloud sandbox, VM, and agent-runtime providers for a given AI-agent workload, computing monthly cost from each provider's public pricing pages with a quoted source behind every answer.

Open-source local AI agent inference benchmark from Artificial Analysis. It replays recorded agent trajectories against OpenAI-compatible model servers to measure throughput and latency on laptops and workstations.

Open-source runtime for running autonomous AI agents in policy-controlled sandboxes. OpenShell enforces filesystem, process, network, and credential boundaries outside the agent process, with auditable controls and formally checked policy changes.

Free, open-source desktop app for running and training AI models on local hardware, with local LLM, diffusion image/video, MLX, GGUF, audio, agent-tool, and model-serving workflows across macOS, Windows, and Linux.

Open-source continuous-improvement service for AI coding agents that captures Codex and Claude Code sessions, identifies recurring workflow patterns, and turns them into reusable skills, checks, prompts, and recommendations.

Open-source, local-first memory layer for AI coding agents that indexes Claude Code, Codex, pi, and Hermes session traces into a searchable Lance dataset, giving agents provenance-linked recall across sessions and machines.

Open-source agent-tool registry that gives coding agents one credential to discover and call a catalog of SEO, research, social, enrichment, advertising, and scraping endpoints, with provider rates shown and catalog usage billed from a prepaid balance.

On-demand API and tool registry for AI agents that lets them discover, inspect, and run 1,800+ tools across 59+ providers through a shared balance, with pay-per-call billing and Skill, MCP, CLI, and HTTP API integrations.

Open-source, read-only PostgreSQL observability CLI for developers and AI agents. It analyzes native database statistics, reports health and query findings, and exposes deterministic diagnostics through MCP for agent workflows.

AWS-built AI-powered development environment that unifies an agentic IDE, terminal CLI, browser sessions, and mobile monitoring around one shared agent harness, with spec-driven development, parallel agents, MCP, and headless CI/CD workflows.

Open-source, local-first desktop AI coworker that turns outcome-based requests into finished work across files, repositories, Slack, calendars, and connected tools, while asking for approval before consequential actions.

Brainbase's Universal Managed Agents API is a single managed API to deploy coding and knowledge-work agents across 8+ harnesses (Claude Code, Codex, OpenCode, Cursor, and more) with interchangeable models — Brainbase runs the agent loop, sandboxed computers, and credentials for you. One request provisions an agent with its own environment, so platforms ship agent infrastructure without managing it.

pdfcn is a free, MIT-licensed collection of PDF components and blocks for React, built on the Takumi and Forme renderers and installed through the shadcn/ui registry — themes, tables, invoices, and reports you copy into your own codebase. It also ships an MCP server so AI coding tools can discover and compose pdfcn documents.