DEV.co
AI Frameworks · icip-cas

PPTAgent

PPTAgent is an open-source agentic framework for automated PowerPoint generation using LLMs. It supports both template-based and freeform visual design, integrates deep research capabilities, and can generate presentations from text prompts, PDFs, and spreadsheets with optional offline operation.

Source: GitHub — github.com/icip-cas/PPTAgent
4.8k
GitHub stars
564
Forks
Python
Primary language
MIT
License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

FieldValue
Repositoryicip-cas/PPTAgent
Ownericip-cas
Primary languagePython
LicenseMIT — OSI-approved
Stars4.8k
Forks564
Open issues8
Latest releasev2.0.0 (2025-12-16)
Last updated2026-06-29
Sourcehttps://github.com/icip-cas/PPTAgent

What PPTAgent is

Python-based agent framework leveraging LLMs with MCP server support, sandbox tool execution (20+ tools), text-to-image generation, and pluggable model backends. Includes a fine-tuned 9B DeepPresenter model, Docker-based deployment, and CLI/web interfaces.

Quickstart

Get the PPTAgent source

Clone the repository and explore it locally.

terminalbash
git clone https://github.com/icip-cas/PPTAgent.gitcd PPTAgent# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Automated Business Report Generation

Generate Q-series reports, financial summaries, and data-driven presentations from raw data files (XLSX, PDF) with automatic research integration and asset creation.

Research-Driven Presentation Creation

Combine web search (Tavily), PDF parsing (MinerU), and agentic reasoning to auto-create research-backed presentations with minimal manual effort.

Enterprise Content Automation

Deploy via Docker Compose for internal teams needing fast, consistent presentation generation with optional offline mode for sensitive data handling.

Implementation considerations

  • Environment setup on Linux/macOS requires manual dependency preparation (Docker, Node.js, Playwright, poppler); macOS offers semi-automated installation via Homebrew.
  • Fine-tuned DeepPresenter-9B model is recommended for best quality; using generic open-source models may degrade presentation coherence and visual design.
  • Optional services (Tavily, MinerU) significantly improve quality; offline mode available but requires local MinerU deployment; budget for external API costs.
  • Sandbox execution of 20+ tools runs in isolated Docker containers; ensure Docker daemon stability and resource limits for production deployments.
  • Configuration via YAML and JSON files; interactive onboard CLI available to guide initial setup and dependency verification.

When to avoid it — and what to weigh

  • Windows-Only Environments — Project explicitly does not support Windows natively; WSL is required, adding deployment complexity for Windows-centric organizations.
  • Strict Real-Time Latency Requirements — LLM inference, research integration, and asset generation introduce latency unsuitable for interactive, sub-second response scenarios.
  • Minimal External Dependency Tolerance — Deployment requires Docker, Node.js, Playwright, poppler, and optional services (Tavily, MinerU); self-hosted models need significant compute resources.
  • Proprietary Model Lock-In Constraints — Framework recommends fine-tuned DeepPresenter model; compatibility with proprietary or custom LLMs requires integration development.

License & commercial use

MIT License: permissive, allows commercial use, modification, and redistribution with attribution. No patent grants or liability limitations in standard MIT terms.

MIT permits commercial use. However, verify that bundled dependencies (DeepPresenter fine-tuned model, optional services like Tavily and MinerU) have compatible commercial terms. Tavily and MinerU require separate API agreements and may have commercial restrictions. Consult legal review for production commercial deployment.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

SignalAssessment
MaintenanceActive
DocumentationAdequate
License clarityClear
Deployment complexityModerate
DEV.co fitGood
Assessment confidenceHigh
Security considerations

Sandbox tool execution in isolated Docker containers mitigates code injection risks. However, LLM inference can be exploited via prompt injection; sanitize user inputs. External API keys (Tavily, MinerU, LLM providers) must be protected (environment variables, secrets management). Offline mode reduces network exfiltration risk. No audit trail, encryption, or RBAC mentioned. Security posture requires review before handling sensitive data.

Alternatives to consider

Microsoft Copilot for Office / PowerPoint Designer

Proprietary, fully managed, integrated with Office ecosystem, but closed-source, limited customization, and subscription-based.

Beautiful.ai

SaaS-based AI presentation tool with polished UI; no deployment overhead, but vendor lock-in, limited offline capability, and licensing costs.

Gamma.app / Deck.gl

Web-based presentation generators with AI assist; simpler setup but less control, limited research integration, and hosted-only delivery models.

Software development agency

Build on PPTAgent with DEV.co software developers

Start with PPTAgent CLI on macOS/Linux, or deploy via Docker for team use. Review dependencies and optional API services before production rollout.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

Related on DEV.co

Explore the category and the services that help you build with it.

PPTAgent FAQ

Can I use PPTAgent without Docker?
Yes, CLI mode works with local Python environment (via `uv`). Docker is optional but recommended for sandbox tool isolation and team deployments.
What LLM models are supported?
Fine-tuned DeepPresenter-9B is recommended. Framework is pluggable; OpenAI, local models, and other backends via config. Exact support matrix requires codebase review.
Is offline operation fully possible?
Partial: set `offline_mode: true` and deploy MinerU locally. Web search and some cloud services can be disabled. External LLM endpoints still required unless a local model is configured.
What is the cost of running PPTAgent?
Open-source codebase is free. Costs depend on: external APIs (Tavily, MinerU, LLM provider), compute for local model inference (DeepPresenter-9B ~4-5GB VRAM), and Docker/infrastructure. Estimate $0.10–$2.00 per presentation for API-heavy workflows.

Software developers & web developers for hire

Adopting PPTAgent is usually one piece of a larger software development effort. As a software development agency, DEV.co provides software development services and web development expertise — pairing senior software developers and web developers with your team to design, build, and operate ai frameworks software in production.

Ready to Automate Presentations?

Start with PPTAgent CLI on macOS/Linux, or deploy via Docker for team use. Review dependencies and optional API services before production rollout.