---
id: engines/oasis-social-simulation
about:
  label: OASIS (Open Agent Social Interaction Simulations)
  description: >-
    Open-source simulator from CAMEL-AI of social media platforms like X and
    Reddit, populated by up to a million language-model users who post, follow
    and repost under a recommendation algorithm.
  tags:
    - LLM agents
    - social media
    - misinformation
    - polarization
    - recommendation systems
    - CAMEL-AI
tool:
  form: open-source Python package
  group: People and populations
  homepage: 'https://github.com/camel-ai/oasis'
  licence: Apache-2.0
  cost: free software; LLM calls are paid per token
open:
  source: open
  data: open
  api: unknown
  participation: open
  licence: Apache-2.0
  note: >-
    A dataset is published on Hugging Face; documentation at
    docs.oasis.camel-ai.org.
domains:
  - people
relations:
  - predicate: alternative_to
    target: engines/agentsociety
  - predicate: alternative_to
    target: engines/mirofish
provenance:
  origin: agent
  sources:
    - title: camel-ai/oasis (README)
      url: 'https://github.com/camel-ai/oasis'
    - title: >-
        OASIS: Open Agent Social Interaction Simulations with One Million Agents
        (arXiv abstract)
      url: 'https://arxiv.org/abs/2411.11581'
    - title: 666ghj/MiroFish (README)
      url: 'https://github.com/666ghj/MiroFish'
    - title: >-
        camel-ai/oasis (GitHub API repository record: created and last pushed
        dates)
      url: 'https://api.github.com/repos/camel-ai/oasis'
  created: '2026-10-23'
  reviewed: '2026-10-23'
  reviewed_by: 'Claude, from the sources listed'
  status: draft
---
OASIS imitates a social media platform and fills it with simulated users, each driven by a large language model (LLM).[^1] It is built for studying how things spread online: information, group polarization, where opinions in a group drift to extremes, and herd behaviour.[^1] It comes from the open-source CAMEL-AI project, and its 2024 paper has 23 authors, Ziyi Yang first and Jing Shao last.[^1,2]

The simulated platform has the parts that shape real ones.[^1,2] Users have profiles and a follow graph that changes as they act.[^2] Users can take 23 actions, including posting, commenting, reposting, quoting, following, muting and reporting a post.[^1] Each user's feed is chosen by a recommendation system, either by interest or by a "hot score" of popularity.[^1,2] Versions modelled on X and on Reddit are included.[^1,2] The authors report that it reproduces information spreading, polarization and herd effects on both, and that larger simulated crowds produced stronger group dynamics and more varied opinions.[^2]

You install it with pip as camel-oasis, supply a file of user profiles and a model, and step the simulation forward in Python; results go to a database.[^1] It works with OpenAI models and others through the CAMEL framework.[^1] The README gives a cost reference: 100 users for one step used about 336,000 input tokens on Alibaba's Qwen-turbo model.[^1] Cost grows with users, steps and the share of users active each step, so a million-user run is a large undertaking.[^1] Since 2025 a researcher can also interview a simulated user mid-run.[^1]

The README lists follow-on research built on OASIS, on collusion, fraud and marketplaces,[^1] and [MiroFish](/engines/mirofish/) uses it as its simulation layer.[^3] The repository was created in November 2024 and was still being updated in October 2026.[^4] Each study has to test for itself how closely the simulated users behave like people on a real platform. Licence Apache-2.0.[^1]