Beale-CC
Local MCP server that bridges Beale research workspaces to Claude Code, providing typed memory, workspace search, revisioned runbooks, and research profile context via a stdio plugin.
README
Beale-CC
Beale-CC is a local, agent-neutral compatibility bridge for Beale research workspaces. It provides a bundled Claude Code plugin backed by a stdio MCP server that serves MCP 2026-07-28 while retaining the legacy handshake required by older Claude Code clients.
The bridge uses Honeycrisp's public workspace-tool API rather than Beale's Electron IPC or direct SQL writes. A Claude Code session and Beale can therefore continue the same research through typed memory, workspace-default memory search, subject-explicit search, revisioned runbooks, the active research profile, and recorded workspace context.
It does not contribute a system prompt, prompts, skills, commands, hooks, subagents, AGENTS.md, CLAUDE.md, or delegation policy. Those remain owned by the agent using the MCP server.
Requirements
- Node.js 22.19 or later
- A sibling
honeycrispcheckout when developing or rebuilding Beale-CC - Claude Code for plugin installation
Build and test
npm ci
npm test
npm run validate:plugin
The build produces a standalone ESM plugin server at plugins/beale-cc/dist/server.mjs. The installed Claude plugin is copied into Claude Code's cache, so the bundle includes every runtime dependency and does not reach back into this source checkout.
Install in Claude Code
claude plugin marketplace add C:\path\to\Beale-CC
claude plugin install beale-cc@beale-cc-local
Open Claude Code in the research workspace. The plugin passes ${CLAUDE_PROJECT_DIR} to the server at startup, permanently binding that server process to the active workspace.
After rebuilding an already installed plugin, update the marketplace and reinstall the plugin so Claude Code refreshes its cached copy.
MCP surface
Tools:
workspace_contextmemory_search,memory_get,memory_save,memory_correct,memory_linkrunbook_list,runbook_get,runbook_create,runbook_append
Resources:
beale://workspace/contextbeale://workspace/profilebeale://profilesbeale://profile/{id}beale://memory/catalogbeale://memory/recentbeale://memory/{id}beale://runbooksbeale://runbook/{id}
Profile capabilities can disable the corresponding memory or runbook surface. beale://workspace/profile is the active, snapshot-pinned workspace profile. beale://profiles exposes the built-in Security Research and Mathematics catalogs for discovery, while beale://profile/{id} exposes one catalog. These resources omit agent prompts and model-job configuration and do not change the active workspace profile. Select a profile through Beale or the workspace's normalized .honeycrisp/profile.json; this MCP server does not add model-facing profile mutation.
Runbooks are recorded procedures and are never executed by this server.
Workspace continuity
Canonical durable storage remains Honeycrisp-owned:
- Global memory database:
~/.honeycrisp/memory.sqlite - Global artifact store:
~/.honeycrisp/artifacts - Optional workspace profile:
.honeycrisp/profile.json - Beale workspace metadata, when Beale opens the directory:
.beale/
Workspace identity is the shared deterministic ID derived from the resolved absolute workspace path. Beale-CC prefers Beale's durable workspace_research_subjects binding and active immutable profile snapshot when present. For a Claude-first workspace, Honeycrisp creates the same workspace ID and a workspace-placeholder subject; Beale retains that subject during later onboarding when it already owns durable memory. Workspace copies or moves have a new path-derived identity.
Memory search defaults to the current workspace. Subject-wide recall requires scope: "subject"; there is no global cross-workspace search.
External Claude Code sessions do not create Beale run or transcript rows. Continuity covers durable research state and workspace guidance, not Beale session history.
Restart the Claude Code MCP server after Beale changes the workspace subject, authorization metadata, or active research profile. Beale-CC refuses tool and resource access when it detects that the live process's subject binding has changed, preventing writes to a stale subject.
Security boundary
- The server is local stdio only and opens no network listener.
- The workspace root is fixed at process startup and cannot be supplied by a model tool call.
- Database paths, credential material, registry paths, and the unscoped global artifact manifest are not exposed through MCP results.
- Recorded authorization is projected as context; reading it does not grant authorization or enable live-target testing.
- Beale-CC, Beale, Honeycrisp, target code, and research tools run with the current user's host privileges. Use an externally launched VM or container when OS isolation is required.
- Memory and runbook content is untrusted research data. Honeycrisp remains authoritative for taxonomy validation, evidence references, revisions, and workspace ownership.
The server implementation follows the MCP 2026-07-28 specification and the official TypeScript SDK stdio serving guidance. Plugin packaging follows the Claude Code plugin reference.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.