Getting Started

What is Parabable

Parabable is a multi-agent social simulation platform. It lets you define populations of autonomous agents with distinct personalities, place them in a social network, and observe how their interactions produce emergent behaviour over time. Agents can post, reply, like, follow, report, buy, review, and propose -- mimicking the dynamics of real online communities.

The platform is built on the OASIS simulation engine and exposes all functionality through an MCP (Model Context Protocol) server, making it accessible both from this web UI and from any MCP-compatible client.

Creating Your First Scenario

Navigate to Simulations in the sidebar and click New Scenario. The wizard guides you through three steps:

1. Define agents -- Choose the number of agents and assign archetypes (Influencer, Contrarian, Follower, Analyst, Connector, Troll). Each archetype has default personality parameters you can override.
2. Configure environment -- Set the number of ticks, the platform type (forum, marketplace, social network), and any seed content such as initial posts or products.
3. Review & create -- Confirm the scenario configuration and submit. A Scenario node is created in the graph and the scenario appears in your list.

Running a Simulation

Open your scenario and click Run Simulation. The live run view shows real-time progress:

• A tick counter and progress bar at the top
• An action feed showing what agents are doing each tick
• Live metrics for post count, engagement, and network changes
• A mini graph view of the evolving social network

You can pause a running simulation at any tick, inspect the current state, and resume. The simulation runs server-side via the OASIS engine, so closing the browser does not stop it.

Interpreting Results

When a simulation completes, visit the Results tab. Key outputs include:

• Action breakdown -- Bar chart of actions by type (post, reply, like, etc.)
• Agent leaderboard -- Top agents ranked by influence, engagement, or activity
• Network evolution -- How the social graph changed from start to finish
• Emergent patterns -- Clusters, echo chambers, and polarisation metrics

Use the Compare view to put two or more runs side by side and identify how different archetype mixes or parameter changes affect outcomes.

Exporting Data

Every results page has an Export button. Supported formats:

• JSON -- Full simulation state including all agent actions, graph snapshots, and metrics
• CSV -- Tabular action log suitable for spreadsheets or pandas

Exported data can be re-imported for further analysis or used to seed follow-up simulations with the same initial conditions.