Sparkles is a collaborative development platform that lets entire teams, not just engineers, direct AI coding agents on existing GitHub repositories. Agents work in isolated cloud sandboxes and propose changes as pull requests, with every session streamed live and visible to the whole organization. The platform supports multiple agent runtimes including Claude Code, Codex, and OpenCode, and integrates with 3,000+ tools like Slack, Linear, and Sentry.
Funding
Funding not disclosed
Founders
Product
Problem
Software development has traditionally been gated behind engineering expertise, leaving non-technical team members unable to contribute directly to code changes. Existing AI coding tools are either built for individual developers in their own IDEs or focus on creating new apps from scratch, rather than improving the repositories teams already rely on daily.
Solution
Sparkles provides a shared online development environment where anyone on a team can direct AI coding agents to work on existing GitHub repositories. Agents operate in isolated, disposable cloud sandboxes and return all changes as pull requests, so existing branch protections, code reviews, and CI pipelines remain fully intact. Team members can start a run from the web app or directly from Slack, then watch every command, file edit, and tool call stream live in real time. The platform supports multiple agent runtimes—Claude Code, Codex, and OpenCode—and automatically connects 3,000+ pre-integrated tools like Linear, Sentry, and Doppler to every run without per-session setup.
Target Audience
Primary users are software teams at startups and enterprises that want to involve non-engineers in shipping code changes, including product managers, designers, and support staff, alongside engineers who need a shared, reviewable workflow for AI-assisted development on existing repositories.
Features
- Multi-agent support for Claude Code, Codex, and OpenCode with per-run model selection and credit consumption based on actual inference cost
- Isolated cloud sandboxes per project that are disposable and cannot affect production, with changes only proposed as GitHub pull requests
- Live streaming workspace showing every command, file edit, and tool call in real time, with the ability to jump in mid-run with follow-up instructions
- Slack integration allowing team members to @mention Sparkles in channels to start runs and receive PR updates without leaving their chat interface
- 3,000+ pre-connected integrations including GitHub, Slack, Linear, Notion, Sentry, Doppler, Infisical, Jira, Figma, and Stripe, shared across the entire organization
- Enterprise deployment options including VPC or self-hosted infrastructure, SAML SSO, SCIM provisioning, and custom security review
- Sandbox API with bearer-key authentication, stream event contracts, and OpenAPI reference for programmatic control of managed coding environments
- CLI tool for running local agents that can mirror sessions to the Sparkles dashboard or hand off to cloud sandboxes