Pacific Slate is a personal AI system designed to operate over an individual's own knowledge and followed sources, performing useful work in the background rather than only responding to direct queries. It is structured to run on a private remote server, maintaining continuity by remembering context across sessions and reasoning over a private, indexed knowledge base. The system is composed of specialized agents managed by an operator, who routes tasks to the most suitable agent for research, coding, analysis, review, and other functions, each agent backed by a model chosen for its appropriateness and cost. Automatic fallback across different model vendors ensures reliability, and the architecture allows for easy swapping of models via configuration changes.
A key feature of Pacific Slate is its proactive routines: it continuously ingests data from wired-in sources such as mail, calendar, messages, and various feeds or accounts, keeping itself updated on the user's environment. These routines synthesize information from multiple sources and deliver concise briefs, filtering out noise and logging discarded information. The tool layer consists of approximately two dozen tool-servers accessible to agents through the Model Context Protocol (MCP), supporting functions like memory, knowledge retrieval, code execution, web search, and infrastructure management. New tools can be added by registering a server, and agents dynamically locate the right tool by searching an index.
The user interacts with Pacific Slate through a standard Claude client over an MCP connection, enabling access to the system's capabilities via familiar interfaces. The canvas interface presents results as live, movable cards, each tagged with the model used and the cost of the call, allowing for transparency and traceability. Every model call is metered against a budget and fully traced, with spending checked before execution to enforce limits. The system also includes self-maintenance routines that monitor dependencies, flag and apply safe updates, and handle routine fixes, escalating issues that require human judgment.
Privacy is a central design principle: Pacific Slate is built so that the user's data, memory, and tools remain private and under their control, with models accessed as external, swappable components. The architecture emphasizes that the system is private by design, avoiding vendor lock-in and ensuring that personal history and data are not owned or retained by third-party providers.
Pacific Slate is an Autonomous agents & workflows product. It focuses on running a private, customizable AI assistant that integrates multiple models and agents for personal automation. Pacific Slate is an open-source project aimed at technical users seeking private AI assistants. The project is open source (MIT). It runs on the web, API, and the command line, and it can be self-hosted.
It is developed by Pacific Slate Maintainers, and the product first shipped in 2026. The project is developed in the open on GitHub with 4 commits in the last 90 days. Among its 5 catalogued features are multi-agent orchestration, self-hosted deployment, and model agnostic. It exposes integrations via an MCP server and a public API.
Latest indexed changes and source events
Show HN: Pacific Slate – a self-hosted, model-agnostic multi-agent AI assistant verified by the PulseGate indexer
Other apps tracked under the same category.