Giselle

Giselle is a visual builder for AI agents and multi agent workflows. Agents are assembled on a canvas by connecting model nodes, data sources and tools rather than by writing code or hand tuning prompts, then deployed to managed cloud hosting or run on a team's own infrastructure. The emphasis is on chaining several reasoning steps and several models together so that a single request produces a finished artifact, such as a specification or a review, rather than a chat reply. It is aimed at small product and engineering teams that want routine work automated without adding headcount. Prebuilt agents cover deep research across web and internal sources, automated pull request review informed by commit history and coding standards, generation of product requirement documents and execution plans from repository activity and database content, and keeping readme files, release notes and published content in sync when code merges. Giselle connects to GitHub, to PostgreSQL databases and to document and repository vector stores, and routes each task to a suitable model from a catalog of more than fifteen across Anthropic, Google and OpenAI. The vendor states the platform is ISO/IEC 27001 compliant with SOC 2 in progress, and the codebase is published on GitHub for self hosted use. The open source edition is free. A hosted Free plan is also free and includes 30 minutes of AI model usage, basic models, managed cloud deployment and a single user seat. Pro is $20 per month for one seat and adds $20 of AI credits each month, premium models, data and vector stores and email support, with usage beyond the included credits charged at a 10 percent markup on base rates. A Team plan at $100 per month for up to 10 users with collaboration and priority support is listed as coming soon.