You type a slash command, /guepard, into your AI code editor. The agent, aware of the project’s sprawling Postgres instance, needs to test a new data transformation. In under six seconds, a full, isolated clone of the production database appears, a disposable sandbox that will vanish when the agent’s work is done. This is the workflow Guepard is selling, not just to developers, but to the AI agents that are beginning to work alongside them.
It’s a vision that turns the database from a monolithic, carefully guarded artifact into something fluid and forkable. The San Francisco-based startup, which emerged from stealth in September 2025 with a $2,100,000 pre-seed round, calls its platform "serverless" and pitches it as "Git for databases" [Newsfile, September 2025].
The Wedge of Instantaneous Experimentation
Guepard’s core bet is that the friction of data provisioning has become a critical bottleneck. Guepard’s platform, which connects to source databases like Postgres and MySQL in read-only mode, uses incremental, deduplicated snapshots to spin up what it calls "production-identical" environments in seconds [guepard.run].
By offering a Model Context Protocol (MCP) interface, Guepard allows tools like Cursor or Claude to securely query and operate on these versioned data branches [Perplexity Sonar Pro Brief].
Why Schematic and Skydeck Wrote the Check
The $2,100,000 pre-seed, led by Schematic Ventures with participation from the UC Berkeley Skydeck Fund, arrived as Guepard publicly launched its self-service platform [The Globe and Mail, September 2025]. Early claims from the company suggest the approach can cut physical resource requirements by an estimated 50% and minimize data access wait times [StartupSeeker].
The team is led by CEO Koutheir Cherni and co-founders Ghassen Menaouar and Hani Chalouati [Bizprofile.net]. With a headcount estimated between 2 and 10 employees [Prospectoo, retrieved 2026], the focus is squarely on product execution.
Navigating a Field of Data Contenders
| Competitor | Primary Focus | Key Differentiation vs. Guepard |
|---|---|---|
| Neon | Serverless Postgres | Focuses on compute autoscaling and branching for the Postgres ecosystem specifically. |
| Dolt | Version-controlled SQL database | Embeds Git-like versioning directly into the database engine itself. |
| Supabase | Open-source Firebase alternative | Offers database branching as part of a broader suite of backend-as-a-service tools. |
The Next Twelve Months of Proof
- Landing lighthouse customers.
- Quantifying the developer love.
- Deepening the AI agent workflow.
The success of that bet hinges on whether teams are ready to work with data as casually as they now work with code.