The quantitative workbench for
prediction market trading.

Your AI analyst, embedded in the notebook

Sharpe puts a reasoning AI assistant directly inside your Jupyter workflow. Describe what you need in plain language and it plans multi-step analyses, writes Python code, executes cells, and iterates on the results -- all in your notebook.

It is not a chatbot bolted onto an IDE. It reads your existing notebook state, understands your data schema, and proposes the next analytical step. Approve it, edit it, or let it run autonomously.

Capabilities

• Multi-step analysis planning with automatic dependency resolution
• Code generation calibrated to your existing imports and style
• Iterative refinement: if a cell errors, the agent debugs and retries
• Approval gates for sensitive operations like writes and API calls

Native Jupyter integration

Sharpe runs on real JupyterHub infrastructure with real-time notebook sync via Yjs. Every cell execution happens on a real kernel with full access to your Python environment -- no sandboxed previews.

Your notebooks are standard .ipynb files. Export them, version them in Git, open them in any Jupyter client. No lock-in.

Details

• Real-time collaborative editing with Yjs conflict resolution
• Persistent kernels with configurable environments and packages
• Cell execution with streaming stdout, stderr, and rich outputs
• Full notebook lifecycle: create, clone, snapshot, and restore

Connect your organization's knowledge

The AI assistant is only as useful as the context it can access. Sharpe lets you connect Confluence, Dremio, internal wikis, and other data sources so the analyst understands your domain.

When you reference a document with @, the assistant pulls in the relevant context automatically. No copy-pasting. No stale screenshots of dashboards.

Integrations

• Confluence pages and spaces with incremental sync
• Dremio datasets with schema introspection
• Jupyter notebooks as referenceable knowledge
• Extensible connector framework for internal tools

Real-time collaboration

Share notebooks with your team and edit simultaneously. Sharpe handles conflict resolution so you never lose work, even when multiple people edit the same cell.

Conversations with the AI analyst are shared too. Your teammate can pick up an analysis thread, add context, and keep the investigation moving without starting over.

Features

• Simultaneous editing with cursor presence and awareness
• Shared AI conversation threads across team members
• Team workspaces with shared connectors and configurations
• Granular permissions: viewer, editor, admin roles

Full audit trail

Every action the AI takes is logged. Every cell it writes, every API it calls, every file it reads. You get a complete audit trail with timestamps, diffs, and approval records.

For regulated environments, Sharpe provides approval gates that require human sign-off before the agent can execute sensitive operations.

Specifics

• OpenTelemetry traces for every AI interaction
• Configurable approval rules for writes, deletes, and API calls
• Usage tracking and cost attribution per user and team
• Export audit logs for compliance and review

Built for quantitative teams

Sharpe is built by a team of engineers and researchers who got tired of context-switching between chat interfaces, notebooks, and data catalogs. We wanted one tool that understands the full analytical workflow.

We are based in New York and backed by investors who believe AI-assisted analysis is the future of quantitative research.