# AppWorld

> Benchmark and sandbox of 9 simulated day-to-day apps operable via 457 APIs and populated by ~100 simulated people, for evaluating function-calling and interactive coding agents.

- id: `appworld`
- category: Information (`information`)
- url: https://appworld.dev/
- repo: https://github.com/StonyBrookNLP/appworld
- docs: https://github.com/StonyBrookNLP/appworld#readme
- status: active
- pricing: open-source
- license: Apache-2.0
- tags: benchmark, function-calling, coding-agent, agent-evaluation, leaderboard
- related: τ-bench (tau2-bench) (https://indexagentica.com/entries/tau-bench/); Berkeley Function Calling Leaderboard (BFCL) (https://indexagentica.com/entries/bfcl/); WebArena (https://indexagentica.com/entries/webarena/); OSWorld (https://indexagentica.com/entries/osworld/)
- sources: https://github.com/StonyBrookNLP/appworld , https://appworld.dev/ , https://arxiv.org/abs/2407.18901
- added: 2026-10-18
- updated: 2026-10-18
- last verified: 2026-10-18
- featured in: Agent benchmarks compared (https://indexagentica.com/compare/agent-benchmarks/)
- maintainer: Stony Brook NLP
- submitted by: agentica-curator
- page: https://indexagentica.com/entries/appworld/
- json: https://indexagentica.com/api/entries/appworld.json

Agents complete natural tasks by writing code that calls the apps' APIs on behalf of a supervisor; Evaluation uses database-state-based unit tests that also catch collateral damage. Includes a task explorer, API explorer, leaderboard and an AppWorld MCP server/client for evaluating terminal agents such as Codex and Gemini CLI. ACL 2024 Best Resource Paper (arXiv 2407.18901). Installs from PyPI as `appworld`.