Strategic Simulation • Multi-Agent AI • Open Source

Simulate Strategy.
Predict Outcomes.

Strategic Interaction Simulation Powered by AI Agents

SparkFish uses multi-agent AI to model how different actors behave in any scenario — business negotiations, market dynamics, geopolitical situations, competitive strategy. Upload a seed document, run a simulation, and get actionable predictions backed by agent-based modeling and LLM reasoning.

⬇ View on GitHub See How It Works
🦠
🔬 CAMEL-OASIS Engine Multi-agent simulation core
👥 Agent Personas Customizable actor behaviors
📈 Prediction Reports AI-generated outcome analysis
🌍 Multi-LLM Ready Claude, GPT, Gemini, and more

From Seed Document to Strategic Insight

SparkFish runs a full agent-based simulation pipeline — from document analysis to entity extraction to multi-round strategic interaction.

1

Upload Seed Document

Feed SparkFish a business plan, geopolitical brief, market analysis, or any strategic document describing the scenario you want to analyze.

2

Entity Extraction

AI extracts key actors, their interests, constraints, and relationships. SparkFish builds a structured map of the strategic landscape.

3

Multi-Agent Simulation

Each actor gets an AI agent persona. Agents interact across multiple rounds — negotiating, competing, cooperating — following their programmed strategies.

4

Prediction Report

SparkFish synthesizes the simulation results into a structured prediction report with probability-weighted outcomes and strategic recommendations.


Built for Strategic Depth

SparkFish combines multi-agent simulation with LLM reasoning to give you insights that single-model analysis can't match.

👥

Multi-Agent Personas

Each actor in the simulation gets a dedicated AI agent with a distinct persona — interests, constraints, behavioral tendencies, and strategic objectives.

Customizable Roles Behavioral Modeling Strategy Assignment
📄

Seed Document Analysis

Upload any strategic document — PDF, text, URL — and SparkFish extracts the relevant actors, relationships, and scenario parameters automatically.

PDF Support URL Fetching Auto-Extraction
📈

Prediction Reports

SparkFish synthesizes the simulation into a structured report with probability-weighted outcomes, key decision points, and strategic recommendations.

Probability Weights Decision Analysis Recommendations
🌐

Deep Interaction Chat

After a simulation, ask SparkFish follow-up questions in natural language. Probe specific outcomes, actor motivations, or alternative scenarios with a conversational layer on top of the structured results.

Conversational Follow-up Scenario Probing Natural Language Q&A

Multi-LLM Support

SparkFish supports Claude, GPT-4, Gemini, and local models via Ollama. Configure the model that fits your use case and budget.

Claude GPT-4 Gemini Ollama
🛡

AGPL-3.0 Open Source

SparkFish is fully open source under AGPL-3.0. Fork it on GitHub, run it locally, or deploy it on your own infrastructure. No vendor lock-in.

AGPL-3.0 Self-Hosted Fork on GitHub

When to Use SparkFish

SparkFish is a strategic analysis tool. It's most powerful when the outcome depends on multiple actors with different goals, constraints, and information.

🏢

Business Negotiations

  • Model how counterparties will react to your proposals
  • Explore "what-if" negotiation scenarios before you enter the room
  • Identify which terms are most likely to close a deal
  • Predict concession patterns and walk-away points
📈

Market Strategy

  • Simulate competitive responses to a new product launch
  • Model pricing dynamics in a market with multiple players
  • Identify which market moves trigger the strongest reactions
  • Explore partnership and merger scenarios
🌍

Geopolitical Analysis

  • Model multi-party diplomatic scenarios
  • Explore how trade policy changes affect alliances
  • Identify likely actor responses to policy moves
  • Stress-test strategic assumptions across multiple rounds

From Upload to Prediction

SparkFish walks through a structured simulation pipeline — each step feeds the next.

What SparkFish produces

After uploading a seed document and configuring actors, SparkFish runs a multi-round simulation and delivers a structured report. Here's what the output contains:

  • Actor Profiles — extracted interests, constraints, strategies, and behavioral tendencies for each simulated party
  • Interaction Log — the full multi-round conversation trace showing how each actor responded and why
  • Outcome Probabilities — probability-weighted predictions for each possible final state
  • Key Decision Points — the moments in the simulation where outcomes were most sensitive to specific moves
  • Strategic Recommendations — AI-generated advice based on the simulation for how to achieve your objectives
⬇ Explore the Code on GitHub
SparkFish — Simulation Report
Actors Detected
Acme Corp — incumbent, risk-averse, defending market share
StartUp Inc — aggressive growth, limited resources, technology advantage
Regulator — stability-focused, long decision cycles
Simulation: 12 rounds
Round 4: StartUp launches disruptor product. Acme considers acquisition vs. counter-launch...
Round 7: Regulator signals review of market concentration...
Round 10: Acme acquires StartUp. Deal closes under regulatory scrutiny.
✓ Most likely outcome: Acquisition at 68% probability
Key Decision Point
Round 4 — Acme's launch decision
Counter-launch delays StartUp growth but invites regulatory review. Acquisition at Round 4 had 2x better value than waiting.
⚠ Regulator involvement increases 3x if market share > 45%
Strategic Recommendation
Approach StartUp for early acquisition before regulatory signals emerge. Target valuation window: before product launch metrics are public.

Built on Proven Multi-Agent AI

SparkFish combines the CAMEL multi-agent framework with the OASIS simulation system, powered by LLMs for realistic actor behavior.

CAMEL-OASIS Architecture

SparkFish is built on a proven multi-agent AI architecture: CAMEL handles role-playing agent coordination, OASIS handles the structured simulation loop, and LLM backends provide the reasoning for each actor.

  • CAMEL role-playing agent framework
  • OASIS structured simulation engine
  • LLM-powered actor reasoning
  • Zep Cloud for entity extraction and memory
  • Docker-based deployment

Self-Hosted & Controllable

Run SparkFish on your own infrastructure. Bring your own API keys. The simulation logic is transparent and auditable — no black-box cloud service.

  • AGPL-3.0 open source — full transparency
  • Self-hosted Docker deployment
  • Bring your own LLM API keys
  • No data leaves your infrastructure
  • Configurable simulation parameters

Ready to stress-test
your strategy?

SparkFish is open source and self-hosted. Fork it on GitHub, run it locally, and start modeling your strategic scenarios today.