- Introduction
Akasha is an enterprise knowledge and memory workspace for humans and agents.
It helps organizations turn scattered work context and experience into knowledge that can be discovered, connected, reused, and grounded in its sources. Teams collaborate in shared spaces, while agents retrieve and work with organizational knowledge within the same permission boundaries.
Akasha brings together a collaborative Wiki, AI-powered knowledge compilation and retrieval, relationship-aware navigation, and MCP-based agent access in a self-hosted platform.
Akasha combines the familiarity of a collaborative Wiki with an AI-ready knowledge layer.
-
π§ Knowledge that stays connected β Pages, spaces, attachments, and compiled knowledge are connected through links, citations, and relationships.
-
π Answers grounded in sources β AI retrieval and answers can point back to source pages and supporting evidence instead of returning unsupported summaries.
-
π€ Built for humans and agents β People and agents access the same knowledge surface within the same workspace and permission model.
-
πΈοΈ Context beyond keyword search β Relationship-aware navigation helps users explore how pages, concepts, entities, and sources are connected.
-
π Self-hosted by design β Organizations can run Akasha in their own environment and control their data, storage, and model endpoints.
-
π Collaborative Knowledge Workspace
Create and organize pages in shared spaces with rich-text editing, Markdown support, attachments, comments, version history, real-time collaboration, and access control.
-
π AI Knowledge Compilation
Queue selected pages and spaces for asynchronous compilation into structured knowledge artifacts. Jobs begin processing after they are enqueued; the pipeline can identify entities, concepts, claims, relations, comparisons, contradictions, and supporting evidence.
-
π Source-Grounded Retrieval and Q&A
Search and ask questions over organizational knowledge using lexical and vector retrieval. When workspace knowledge is used, answers include source-page citations and supporting evidence, with permission-aware results.
-
πΈοΈ Relationship Graph
Explore direct page links and semantic relationships through a visual graph. Graph data is filtered according to the user's access permissions.
-
π€ Agent Access through MCP
Connect agents to Akasha through the
/mcpendpoint. Agents can query knowledge and perform permitted operations on pages, spaces, comments, attachments, and workspace information using API-key authentication.
Akasha keeps original Wiki pages and imported content as the source layer, then builds a structured knowledge layer from them.
The compilation pipeline analyzes source content and produces knowledge artifacts such as summaries, entities, concepts, claims, relations, comparisons, and contradictions. Each artifact retains its source references and evidence so that compiled knowledge can be traced back to the original content.
Wiki pages / imported content
β
Knowledge compilation
β
Structured artifacts + evidence + indexes
β
Retrieval / Q&A / graph navigation
Compilation augments the original Wiki; it does not replace it. Source pages remain available for reading, editing, permission checks, and citation.
Akasha distinguishes between source content and derived knowledge.
Original Wiki pages and imported content remain the primary sources. Compiled artifacts and AI answers are derived from those sources and retain citations, source references, or supporting evidence whenever available.
This makes it possible to:
- trace a compiled claim back to its source page;
- inspect the evidence behind an answer;
- respect source-page permissions during retrieval;
- identify knowledge that needs to be refreshed after its sources change.
When the available evidence is insufficient, the system can indicate that limitation instead of presenting an unsupported conclusion as fact.
Akasha does not treat knowledge as a collection of isolated pages.
The knowledge layer records direct page links and semantic relationships discovered during compilation. These relationships connect pages, sections, entities, concepts, and shared sources, making it easier to explore related context and navigate across a knowledge space.
The relationship graph is an aid for discovery and retrieval, not a replacement for the original source pages. Graph results are filtered according to the user's access permissions.
Akasha is designed for both people and agents.
People use the Wiki interface to create, edit, organize, and discuss knowledge. Agents connect through MCP to search knowledge and perform permitted operations on pages, spaces, comments, attachments, and workspace information.
Both access paths use workspace and resource-level authorization. Agent requests are authenticated with API keys, and supported knowledge queries and operations are recorded for auditing.
The following ideas describe Akasha's product direction. They are exploratory plans and should not be read as a list of features guaranteed to be available in the current release.
Akasha aims to evolve from a knowledge workspace into a broader organizational memory system:
- Factual memory β what happened, supported by source artifacts and provenance;
- Interaction memory β why decisions, disagreements, and trade-offs mattered;
- Action memory β what actions, workflows, and safeguards should follow.
Akasha may eventually allow reusable agent skills, operating patterns, and execution feedback to accumulate across tasks and agents, so that organizational experience becomes easier to reuse over time.
git clone https://github.com/chaterm/Akasha.git
cd Akasha
pnpm installThis repository is a pnpm workspace monorepo. Use pnpm for dependency installation and scripts.
Create the local environment file:
cp .env.example .envGenerate a local application secret and set it as APP_SECRET in .env:
openssl rand -hex 32Do not commit .env or any production credentials to the repository.
Start PostgreSQL with pgvector using the included Compose service, and provide Redis separately:
docker compose up -d db
docker run -d --name akasha-redis -p 6379:6379 redis:7Then apply migrations and start the development servers:
pnpm --filter ./apps/server run migration:latest
pnpm run devOpen http://localhost:3000. For environment requirements, model configuration, service details, and troubleshooting, see docs/development.md.
pnpm run build # Build all workspace projects
pnpm run client:build # Build the frontend
pnpm run server:build # Build the backendBuild artifacts are generated under the corresponding apps/*/dist and packages/*/dist directories.
Akasha provides an MCP endpoint for agents to access the knowledge workspace.
Configure the MCP server with:
- the absolute URL of the deployed Akasha instance followed by
/mcp; - an API key with the required workspace permissions.
The MCP integration supports knowledge queries and permitted operations on pages, spaces, comments, attachments, and workspace information. Requests follow Akasha's authorization rules.
See akasha-plugin/README.md for installation instructions and host-specific configuration examples.
Akasha is designed to run in self-hosted environments. Organizations can control where application data, file storage, and AI model endpoints are configured, while applying their own access-control and operational policies.
Akasha builds upon excellent open-source projects. We gratefully acknowledge:
- Docmost β the collaborative Wiki foundation that the workspace and editor layers build on.
Thank you for your contribution! Please refer to the Contribution Guide for more information.