Catalog Catalogs Agent Frameworks & Harnesses

Agent Frameworks & Harnesses

Frameworks, harnesses, IDEs and runtimes that agents are built and run in. For builders choosing an agent stack.

Code-first frameworks, multi-agent orchestration, coding IDEs and harnesses, computer-use and voice agents, sandboxes, identity and auth, MCP gateways and durable execution, compared on language, hosting, licence, protocol support and lifecycle.

238 systems · 10 sections · 43 acquired, archived, stale or shut down · data as of 2026-10-31

Browse the table → JSON · CSV

Decide

Most referenced

Lowest tier first, then by how many other systems in the catalog reference each one.

  1. Model Context Protocol (MCP spec) T1 Open protocol for LLM-tool I/O
  2. LangChain (framework) T1 LLM orchestration framework
  3. Vercel AI SDK T1 TypeScript LLM toolkit
  4. LlamaIndex (framework) T1 Data framework for LLM apps
  5. OpenAI Agents SDK T1 OpenAI's official agent framework
  6. LangGraph (orchestration) T1 Stateful agent graph runtime
  7. CrewAI T1 Multi-agent orchestration
  8. Claude Code (Anthropic) T1 Terminal-native agentic coding harness
  9. Cursor (IDE) T1 Acquired Agentic IDE; Agent Mode + Background Agents (VS Code fork)
  10. OpenAI Codex CLI T1 Deprecated Terminal-native agentic coding harness (OpenAI)
  11. AutoGen / AG2 T1 Multi-agent conversation framework
  12. n8n (AI agent flows) T1 Workflow automation + AI nodes

All 238 in the table →

Sections

Lifecycle

  • Active 184
  • Maintenance 1
  • Renamed 10
  • Acquired 17
  • Stale 2
  • Inactive 2
  • Deprecated 9
  • Archived 10
  • Shut down 3

Tiers: T1 61 · T2 79 · T3 96 · T4 2 · T5 0

Recently changed

Everything that changed →

Research

Data and tools

  • JSON · CSV, flattened, one row per system, CC-BY-4.0
  • MCP tools: plan_stack, choose_framework_or_store, check_stack, check_compatibility, find_alternatives
  • CLI: landscape stack · landscape choose-stack framework · landscape check <manifest>
  • Data & API · Submit a system