Palam researches the web, creates professional documents, automates browsers, writes and executes code, schedules tasks, and remembers everything across sessions. Open source. Runs on your hardware.
Not a chatbot. An agent that takes action — researching, creating, automating, and learning across every session.
Searches the web, reads and synthesizes sources, follows citations, and produces structured research with inline references. Handles multi-step investigations autonomously.
Generates Word documents, PowerPoint decks, Excel workbooks, and PDFs with professional formatting. Template-driven, publication-quality output.
Controls a real browser via Playwright. Navigates pages, fills forms, clicks buttons, extracts data, takes screenshots. Accessibility-tree-first.
Writes and runs code in a sandboxed environment. Python, JavaScript, Bash. Installs packages, processes data, builds tools, deploys sites.
Semantic memory that survives across sessions. Remembers project context, user preferences, prior research, and past decisions. Every conversation builds on the last.
Connect to Gmail, Slack, Notion, GitHub, Google Calendar, and hundreds more through the Model Context Protocol. Each runs as a self-hosted MCP server.
Palam classifies your intent, loads domain-specific expertise, executes with the right tools, and stores what it learned for next time.
Natural language. "Research competitor pricing and put it in a spreadsheet." "Draft a contract review memo." "Scrape this site and summarize the findings."
Intent classification routes your task to the matching domain skill. Each skill injects deep workflow knowledge, tool configurations, and quality standards for that discipline.
Specialized agents handle the work — research agents search and synthesize, asset agents create documents, coding agents write and test code. Multi-agent orchestration coordinates complex tasks.
Results, decisions, and context are stored to semantic memory. Next session, Palam picks up where it left off — no re-explaining, no lost context.
The capabilities of commercial AI platforms, without the subscription, data collection, or vendor lock-in.
| Palam | Commercial AI | |
|---|---|---|
| Self-hosted & private | Yes | No |
| Open source | MIT License | Proprietary |
| Browser automation | Playwright | Varies |
| Document generation | DOCX, PPTX, XLSX, PDF | Limited |
| Custom skills / extensible | 68 skills + SDK | Closed |
| Persistent memory | Semantic + structured | Basic |
| 400+ integrations | MCP protocol | Varies |
| Monthly cost | $0 | $20-200/mo |
| Your data stays yours | Always | Terms apply |
Each skill injects deep workflow knowledge into the agent at runtime. 15 ship open source. Build your own with the Skill SDK.
Python 3.12+ core with LangGraph orchestration, PydanticAI agents, and pluggable LLM inference. Everything is modular and self-hosted.
Clone, install, run. Or use Docker Compose for the full stack with a single command.
No subscription. No data collection. No vendor lock-in. An AI system that works, running on your hardware, under your control.