What is DocsMint?
DocsMint is an AI-native knowledge workspace designed around a simple idea: the knowledge used by people and AI agents should live in the same place.
Traditional document editors are built primarily for humans. AI knowledge systems often solve the opposite problem and turn documents into an invisible retrieval layer that is difficult for people to inspect or maintain.
DocsMint combines both.
Users can create, edit, organize, search, and share documents through a regular web interface, while applications and AI agents can work with the same knowledge through programmatic interfaces.
Instead of maintaining one knowledge base for a team, another for an AI assistant, and additional copies inside different applications, DocsMint provides a shared source of knowledge that remains visible and editable.
Writing and Knowledge Management
DocsMint provides a visual editor alongside raw Markdown editing.
Documents are stored using a structured editor model, while Markdown remains a convenient format for editing, importing, exporting, and working with developer tools.
The workspace supports the everyday building blocks of a knowledge system:
This makes DocsMint useful for much more than traditional documentation.
A workspace can contain product specifications, research, contracts, internal procedures, meeting notes, technical documentation, articles, proposals, project knowledge, or long-term memory used by AI agents.
The goal is not to force users into a special “AI interface.” People can work with documents normally while the underlying knowledge remains usable by machines.
- documents;
- categories and folders;
- metadata and tags;
- sharing and access control;
- reusable structured content;
- search across the workspace.
Search, Semantic Retrieval and GraphRAG
Keyword search alone becomes increasingly limiting as a knowledge base grows.
DocsMint combines several retrieval methods to find information by both wording and meaning.
Its search layer includes exact matching, lexical search, fuzzy matching, vector retrieval, multilingual expansion, and graph-based retrieval.
Results from these systems are combined to produce a more useful ranked result set.
Documents are automatically chunked and embedded for semantic retrieval. When content or relevant metadata changes, DocsMint updates the affected search representation incrementally instead of rebuilding the entire knowledge base.
GraphRAG adds another layer by connecting related entities and documents, allowing retrieval to move beyond isolated text fragments and use relationships inside the knowledge base.
The result is a workspace that can be searched by humans while simultaneously serving as retrieval infrastructure for AI systems.
Built for AI Agents
DocsMint is designed to be used directly by software, not only through its web interface.
The same workspace can be accessed through:
This allows AI assistants and autonomous agents to read, search, create, and update knowledge without maintaining a separate vector database or private copy of project context.
For example, a coding agent can retrieve technical decisions and product specifications before implementing a feature. A research agent can store findings back into the workspace. Another assistant can later continue the task using the same knowledge.
Because the source documents remain visible to users, knowledge created by agents can also be reviewed and corrected directly.
This makes DocsMint useful as a persistent memory and knowledge layer between different AI tools and applications.
- REST API;
- TypeScript SDK;
- CLI;
- MCP server.
Self-Hosted by Design
DocsMint can run as a complete self-hosted stack on infrastructure controlled by the user.
The deployment includes the application, PostgreSQL, vector search infrastructure, graph data, background processing, and file storage.
A standard Docker-based installation provides the complete workspace without requiring users to assemble a separate collection of knowledge, vector, and agent services.
For AI workloads, deployments can use a hosted model provider or local inference infrastructure.
The self-hosted edition is open source and designed for teams that want control over their documents, embeddings, retrieval infrastructure, and agent-accessible knowledge.
DocsMint is also available as a hosted cloud product for users who do not want to operate the infrastructure themselves.
Use Cases
DocsMint works wherever the same information needs to remain useful to both humans and software.
Store persistent project knowledge that different assistants and agents can retrieve instead of rebuilding context for every conversation.
Keep specifications, architecture decisions, implementation notes, incident reports, API documentation, and release information in one searchable workspace.
Maintain operating procedures, policies, onboarding documentation, research, meeting notes, and internal documentation.
Collect sources, findings, summaries, and related documents while using semantic and graph retrieval to rediscover information later.
Create articles, reports, proposals, contracts, scripts, business documents, and other structured content without separating the writing environment from the knowledge system.
Use DocsMint through REST, SDK, CLI, or MCP as a shared knowledge backend for internal tools and AI-enabled products.
DocsMint started with document creation, but the larger problem turned out to be knowledge continuity.
As AI becomes part of everyday work, documents can no longer exist only as files written for people, while agents maintain separate and increasingly fragmented context.
DocsMint turns the document workspace itself into the shared knowledge layer — readable by people, searchable by applications, and usable by AI agents.