Skip to content
Agent Radar

Agent infra

Hosted agent APIs, security scanners, and the plumbing around agent runtimes.

Latest launch

Agent infra is the layer under the agent: hosted harnesses, sandboxes, and scanners that gate what an agent is allowed to install. If the primary artifact is a library you import, it belongs with frameworks.

Launches

Newest first
  • Phoenix v20.21.0 adds Mistral AI cost tracking

    PhoenixArize Phoenix v20.21.0, released October 11, 2026, adds Mistral AI model pricing to cost tracking and fixes evaluator metadata loss.

  • Cloudflare Sandbox SDK 1.0.0 is a ground-up rewrite

    Cloudflare Sandbox SDKCloudflare released @cloudflare/sandbox 1.0.0 on October 8, 2026, a ground-up rewrite of its Sandbox SDK that requires migrating from 0.x.

  • OpenAI Agents API

    OpenAI AgentsOpenAI's Agents API, in public beta from September 10, 2026, exposes the Codex harness and an optional hosted sandbox.

  • NVIDIA SkillSpector

    SkillSpectorSkillSpector scans agent skills for vulnerabilities and malicious patterns before install. The repository was created March 21, 2026, and includes an MCP server.

Tools in this category

All tools

Cloudflare's SDK for running untrusted or generated code in isolated sandboxes via a Worker.

Daytona is infrastructure for running AI-generated code in isolated sandboxes via SDKs, API and CLI.

DeepEval

Apache-2.0 · Open source

DeepEval is an Apache-2.0 open-source Python framework for evaluating LLM applications and agents.

E2B

Apache-2.0 · Open source

E2B is an open-source sandbox environment for running AI agents in secure, on-demand Linux VMs.

Edi Life OS

MIT · Open source · MCP

Self-hosted life OS for habits, goals, Kanban, finance and focus, with an MCP server for AI.

Langfuse

MIT · Open source · MCP

Open source platform for tracing, evaluating and improving AI agents and LLM applications.

LiteLLM

Open source · MCP

LiteLLM is an open-source library with a unified interface to call 100+ LLMs in the OpenAI format.

Modal runs code in the cloud and offers Sandboxes, secure containers for untrusted user or agent code.

Ollama

MIT · Open source

Ollama is a tool for running open AI models locally or in the cloud, per its own site.

Open WebUI is the platform for running AI on your own terms, connecting to any model, local or cloud.

OpenAI Agents

Open source

OpenAI's agent harness, shipped as a Python SDK update in April 2026 and as the hosted Agents API in September 2026.

OpenRouter is an API with SDKs and an agent package for sending chat requests to models via one endpoint.

Opik

Apache-2.0 · Open source · MCP

Open-source platform for tracing, evaluating and optimizing LLM applications and agents.

Phoenix is an Arize-maintained AI observability and evaluation tool with prompt and dataset tooling.

Portkey

MIT · Open source

Portkey is an AI gateway with guardrails that routes requests to many LLMs through a single API.

promptfoo

MIT · Open source

promptfoo is an open-source CLI and library for evaluating and red-teaming LLM apps.

SkillSpector

Apache-2.0 · Open source · MCP

Apache-2.0 scanner from NVIDIA for agent skills and MCP installs, with a local MCP server that exposes scan_skill.

Widefleet

MIT · Open source

Open-source platform for agent-built apps and workflows with company login, databases and file storage.

Other categories