Chisel Knowledge MCP
The Chisel Knowledge MCP server lets you build and manage knowledge workspaces by ingesting raw content, organizing it into a structured knowledge base, and searching across it.
Workspace Management
List all configured workspaces and inspect their status.
Content Ingestion
Ingest raw text, clipboard content, or a URL into a workspace's inbox.
Inbox Processing
List unprocessed inbox files awaiting compilation.
Obtain deduplication context by searching both inbox and compiled knowledge.
Archive processed inbox files to
inbox/archived/.
Knowledge Base Operations
Write new or updated articles into the knowledge base (updating indexes and logs).
Read the contents of specific knowledge files.
List markdown files in the knowledge directory.
Search across compiled knowledge articles in a workspace.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Chisel Knowledge MCP search second-brain for machine learning resources"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
chisel-knowledge-mcp
Standalone MCP server and library for building and managing knowledge workspaces.
Canonical behavior documentation lives in docs/chisel-knowledge-mcp.md and is routed from docs/CANONICAL_DOCS.md.
Requirements
Node.js 22 or newer
npm
Related MCP server: Notemd MCP Server
Install
npm installLibrary usage
Import the workspace service and related types directly from the package root:
import { WorkspaceService, KnowledgeIndex } from '@teknologika/chisel-knowledge-mcp';The MCP server remains available from the server subpath and through the published binary.
Workspace Workflow
The workspace service and MCP server expose a deterministic inbox pipeline:
knowledge_get_next_inbox_filereturns the first unprocessed inbox file with its content.knowledge_get_dedupe_contextreturns search results from bothknowledge/andinbox/for a file-specific query.knowledge_compile_newwrites a new article intoknowledge/, updatesknowledge/index.md, appendsknowledge/log.md, and archives the source inbox file.knowledge_compile_extendwrites a revised article intoknowledge/, updates the article'sUpdatedentry inknowledge/index.md, appendsknowledge/log.md, and archives the source inbox file.
These tools are deterministic. The LLM that consumes the MCP server decides the article content and the dedupe outcome; the server only performs file and index updates.
Build
npm run buildType check
npx tsc --noEmitConfig
The server reads workspace configuration from:
~/.chisel/config.json
If the file does not exist, the server starts with zero configured workspaces and logs a warning to stderr. See config.example.json for the expected shape.
Claude Desktop
Use this MCP server with Claude Desktop by adding:
{
"mcpServers": {
"chisel-knowledge": {
"command": "npx",
"args": ["-y", "@teknologika/chisel-knowledge-mcp"]
}
}
}Notes
Transport is stdio only.
Logging goes to stderr so stdout stays reserved for MCP protocol messages.
The binary is
chisel-knowledge-mcp.The package root exports the library surface;
@teknologika/chisel-knowledge-mcp/serverresolves to the MCP server entry point.
Available Tools
11 toolsknowledge_archiveC
Move a processed inbox file to inbox/archived/.
Inputs: workspace (required), file (required).
knowledge_ingest_clipboardC
Read clipboard text and write it into a workspace inbox.
Inputs: workspace (required), title.
knowledge_ingest_textC
Write raw text into a workspace inbox.
Inputs: workspace (required), content (required), title.
knowledge_ingest_urlC
Ingest a URL into a workspace inbox.
Inputs: workspace (required), url (required), title.
knowledge_listC
List markdown files in a workspace knowledge directory.
Inputs: workspace (required), directory.
knowledge_list_inboxC
List uncompiled files in a workspace inbox.
Inputs: workspace (required).
knowledge_list_workspacesB
List configured workspaces.
Inputs: no named parameters.
knowledge_readC
Read a knowledge file from a workspace.
Inputs: workspace (required), path (required).
knowledge_searchC
Search workspace knowledge.
Inputs: workspace (required), query (required), limit.
knowledge_workspace_statusC
Inspect a workspace.
Inputs: workspace (required).
knowledge_writeC
Write a compiled article into the workspace knowledge directory.
Inputs: workspace (required), path (required), content (required).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 11 tools
Each tool has a clearly distinct purpose with no overlap: ingest tools handle different input sources (clipboard, text, URL), list tools target different directories (inbox, knowledge, workspaces), and read/write/search/archive/status tools perform unique operations. An agent can easily differentiate them based on their specific actions and targets.
All tools follow a consistent 'knowledge_verb_noun' pattern (e.g., knowledge_ingest_url, knowledge_list_inbox, knowledge_search). This uniform naming convention makes the tool set predictable and easy to navigate, with no deviations in style or structure.
With 11 tools, the server is well-scoped for knowledge management, covering ingestion, organization, querying, and maintenance. Each tool earns its place by addressing a specific need in the workflow, from input handling to file management and search, without being overly sparse or bloated.
The tool set provides complete coverage for knowledge management: ingestion from multiple sources (clipboard, text, URL), listing and reading files, searching content, writing compiled articles, archiving processed items, and workspace inspection. There are no obvious gaps, enabling agents to handle the full lifecycle of knowledge processing.
Maintenance
Related MCP Connectors
Ask your own documents and get answers that cite the source passage. Workspace knowledge bases over MCP for Cursor, Claude and any MCP client.
Personal context for every AI: search, read, and write back to your private Markdown library.
Personal wiki and memory layer for AI assistants. Persistent, structured memory across sessions.
Read-it-later for AI agents. Save links, bookmarks, notes, and videos as Markdown and search them.
Related MCP Servers
- FlicenseAqualityDmaintenanceEnables AI assistants to manage a personal markdown-based knowledge base with natural language interactions. Supports creating, searching, updating, and organizing notes across categories like people, recipes, meetings, and procedures.111-
- AlicenseNot gradedqualityDmaintenanceEnables AI-powered knowledge base management with automated wiki-linking, content generation from titles, web research summarization, and knowledge graph integrity maintenance for Markdown files.3MIT
- FlicenseNot gradedqualityCmaintenanceEnables AI assistants to capture bookmarks, sticky notes, todos, notes, and web archives, and to search and organize personal knowledge and memory across cloud-synced clients.-
- AlicenseNot gradedqualityBmaintenanceEnables building and querying a Markdown-based knowledge graph with hybrid vector, keyword, and graph retrieval, community detection, and evidence-grounded answers through isolated workspaces.218 npm9MIT