Presenton – Open-Source AI Presentation Generator & API
Self-hosted, model-agnostic AI presentation generator with customizable templates, PPTX export, and a programmable API.
Evidence shows: README emphasizes self-hosting, BYOK, local processing, and provides CAN_CHANGE_KEYS environment variable to control API key visibility, indicating some consideration for least privilege. However, no explicit user confirmation mechanism (e.g., confirmation before generation), detailed data flow transparency, specific sensitive data handling policies (e.g., encryption at rest), dependency security audit records, clear external effects (e.g., network requests), rollback mechanisms, or detailed source attribution are found. Therefore, scores are low; deductions due to lack of implementation details and evidence.
Evidence shows: README and test files (e.g., conftest.py, edge_cases) indicate test coverage, and test files demonstrate error handling (e.g., HTTPException 408), suggesting self-consistency and failure messages are somewhat ensured. Dependency locking (uv sync --locked) and Docker image build process indicate dependency availability is considered. However, no comprehensive error handling documentation or failure recovery guide is found, so not full marks.
Evidence shows: README details multiple deployment methods (Docker, Electron, cloud), multiple LLM providers, multiple templates and scenarios (business, education, etc.), indicating broad audience and scenario coverage. Capability boundaries (e.g., supported models, features) are clearly stated, trigger precision (e.g., environment variable configuration) is well documented, and environment fit (multi-platform, multi-architecture) is described. However, no detailed guides for specific user groups or in-depth discussion of edge cases, so not full marks.
Evidence shows: README has clear structure with quickstart, features, deployment config, etc., indicating good information architecture. Installation notes are detailed (Docker, Electron), naming is stable (version numbers, environment variables), examples and FAQ sections exist (e.g., template previews, common questions), known limitations section is missing (not explicitly listed), license is Apache-2.0 and complete, versioning and changelog exist (package.json version, GitHub Actions release workflow), maintenance responsibility is clear (GitHub Actions, test workflows). Deduction: known limitations not explicitly listed.
Evidence shows: README emphasizes output as editable PPTX and PDF, indicating high output usability. Marginal value is evident in self-hosting, no lock-in, multi-provider support, etc. Cost-benefit is addressed through BYOK and local running to reduce costs, but no specific cost analysis is provided. Therefore, score 2; deduction: lack of quantitative cost-benefit analysis.
Evidence shows: README feature claims (e.g., support for multiple providers, export formats) correspond to code and test files (e.g., tests for generation requests), indicating claims are traceable. However, cross-source corroboration is limited as only repository-internal evidence is used. Fact-inference separation is good as README distinguishes feature descriptions from user value propositions. Deduction: lack of external verification.
- Publisher identity is unverified; treat as unknown, do not trust based on brand.
- Static review cannot verify actual runtime behavior; all conclusions are based on source code and documentation.
- Dependency security audit records are not provided; recommend checking dependencies for vulnerabilities before deployment.
- Sensitive data handling (e.g., API key storage) is not detailed; configure with caution.
- Known limitations are not explicitly listed, which may affect expectations in real use.
What does this agent do, and when should you use it?
Presenton is an open-source AI presentation generator that runs as a web app (Docker) or a desktop app (Electron). It creates presentations from prompts or uploaded documents, using either built-in templates or custom designs. It exports fully editable PPTX or PDF files and offers a drag-and-drop editor for manual refinement. Users can bring their own API keys (BYOK) for a wide range of providers including OpenAI, Google Gemini, Vertex AI, Azure OpenAI, Anthropic Claude, and local models via Ollama. It also provides a REST API for generating presentations and a built-in MCP server for model-context-protocol integration. The project emphasizes full control, no SaaS lock-in, and no forced subscriptions.
Presenton generates presentation content from a text prompt or an uploaded document by calling the configured LLM (e.g., OpenAI, Gemini, Ollama). It offers a library of built-in templates (Momentum, Dynamic, Executive, etc.) and renders the content into fully editable .pptx or .pdf files. Users can manually edit slides with a drag-and-drop interface. The tool supports multiple text and image-generation providers, configurable via environment variables. It exposes a REST API endpoint POST /api/v1/ppt/presentation/generate that accepts content, tone, verbosity, slide count, template, and export format, returning a path to the generated file. It also includes an MCP server at /mcp for integration with AI agents.
- A sales manager needs to create a polished pitch deck quickly; they pick the Momentum or Dynamic template and input a few bullet points.
- A privacy-conscious team wants to keep data on-premises; they run Presenton with Ollama and generate presentations locally without sending data to third parties.
- A developer building an internal reporting tool wants to auto-generate slide decks; they call the REST API with a bearer token and receive a downloadable PPTX.
- A designer wants full control over the output; they create custom templates in HTML/Tailwind and use Presenton to generate decks matching brand guidelines.
- A consultant needs to produce a client-ready report in PDF; they select a standard template, set export_as to 'pdf', and download the finished file.
What are this agent's strengths and limitations?
- Fully open-source (Apache 2.0) and self-hostable, ensuring complete data privacy and control.
- Supports a wide range of LLM and image providers, including local models via Ollama, and allows mixing text and image providers.
- Exports fully editable PPTX files and supports custom templates created in HTML/Tailwind or from existing PPTX files.
- Includes a built-in MCP server for easy integration into AI agent workflows.
- Requires users to manage their own API keys for LLM and image providers, which may be daunting for non-technical users.
- Self-hosting demands familiarity with Docker, Node.js, Python, and environment configuration.
- Local model performance (e.g., via Ollama) depends on hardware resources and may produce lower-quality results compared to cloud models.
- Some advanced features like Mem0 memory and LiteParse OCR require additional configuration and may have learning curves.
How do you install or deploy this agent?
You can run Presenton via Docker or as an Electron desktop app. For Docker, pull the image and run: docker run -it --name presenton -p 5001:80 -v "./app_data:/app_data" ghcr.io/presenton/presenton:latest. For Windows PowerShell, use ${PWD}\app_data:/app_data. Then open http://localhost:5001. For the desktop app, clone the repo and in the electron directory run npm run setup:env to install dependencies, then npm run dev to start in development mode, or npm run build:all and npm run dist to build installers.
How do you use this agent?
On first run, follow the setup screen to create the primary administrator account. Then configure your LLM provider by setting environment variables (e.g., LLM=openai and OPENAI_API_KEY) and image provider (e.g., IMAGE_PROVIDER=pexels and PEXELS_API_KEY). You can also change providers in the app settings. To generate a presentation, enter a prompt or upload a document, choose a template and parameters, and click generate. The tool will produce a downloadable PPTX or PDF. For API usage, create an access key under Admin → API keys, then call POST /api/v1/ppt/presentation/generate with the key as a Bearer token and a JSON body (e.g., content, template, export_as).
How does this agent compare with similar options?
Presenton positions itself as an alternative to commercial AI presentation tools like Gamma, Canva, Beautiful AI, and Decktopus. Unlike these SaaS products, Presenton is open-source, self-hosted, and allows users to bring their own API keys, avoiding subscription fees and vendor lock-in.