
Multi-workspace Slack MCP server with draft+confirm safety, vault auto-export, a
io.github.adelaidasofiaPublished by GitHub account adelaidasofiav0.1.1Updated Oct 24, 2026
Multi-workspace Slack MCP server with draft+confirm safety, vault auto-export, and triple-mode…
Overview
A local Slack MCP server that lets an assistant read and search multiple workspaces and send messages through a draft-then-confirm flow.
- What you can do
- List workspaces and run a health check across configured Slack accounts, with tokens redacted in responses and an audit log of every call. Read and search channels, threads, users and messages across workspaces. Draft messages, replies and edits, then confirm or cancel them before anything posts. Add reactions, mark channels read and delete the assistant's own messages immediately. Mirror read channels into local Markdown notes in a vault folder.
- When to use it
- Useful when you work across several Slack workspaces and want an assistant to search and summarize conversations, or to prepare outgoing messages that you review before they are sent.
- Requirements
- Runs locally as a Python process (or via the Claude Code plugin install) with a .env file. Needs Slack tokens per workspace: xoxc plus a cookie, xoxp, or xoxb, with SLACK_XOXC_TOKEN declared as a required secret. Workspace aliases and vault path are set through environment variables. Cookie extraction is manual and takes a few minutes per workspace.
Installation
In SourceWeft
- Open Multi-workspace Slack MCP server with draft+confirm safety, vault auto-export, a in the dashboard and add it to a workspace.
- Enable the server for the chats that should use its tools.
Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.
Other MCP clients
Follow the launch instructions in the repository.
README
slack-mcp
[License] [GitHub stars] [Last commit] [Open issues] [PyPI version] [PyPI downloads] [Built by Mycelium AI]
Multi-workspace Slack MCP server with draft+confirm safety, vault auto-export, and triple-mode auth.
One process, N workspaces, every write goes through draft → confirm. Built because the existing Slack MCP servers are single-workspace per instance and the Anthropic reference impl was deprecated.
Why use this over the alternatives
Install
Open Claude Code, paste:
Then fill in tokens in .env (see SETUP.md for cookie extraction walkthrough).
Legacy install
Register in your project .mcp.json (or via claude mcp add -s user):
Restart Claude Code, then run claude mcp list to verify slack shows up.
Tools (v0.1.1)
Meta: list_workspaces, healthcheck
Read: list_channels, search_channels, read_channel, read_thread, list_users, search_users, get_user_profile, search_messages
Write (draft+confirm): send_message → confirm_send, send_reply_quote → confirm_send, update_own_message → confirm_send, cancel_draft
Write (low-consequence, immediate): add_reaction, mark_read, delete_own_message
Auth modes
Cookie extraction takes ~3 minutes per workspace. See SETUP.md.
Vault auto-export
Every read_channel call mirrors the channel to <vault>/🤖 AI Chats/Slack/<workspace>/<channel>.md. Idempotent (same channel + day overwrites cleanly). Disable via SLACK_MCP_VAULT_EXPORT=false.
Configuration
All config via env vars (loaded from .env at process start):
SLACK_WORKSPACES— comma-separated list of aliases (e.g.onde,mycelium)SLACK_PRIMARY_WORKSPACE— default workspace when tool calls omitworkspace- Per-workspace:
SLACK_WORKSPACE_<ALIAS>_TYPE,_TOKEN,_COOKIE(xoxc only),_TEAM_ID,_LABEL SLACK_MCP_VAULT_PATH— vault root for the auto-export mirror (default~/Documents/Vault)SLACK_MCP_VAULT_EXPORT—true/false(defaulttrue)SLACK_MCP_AUDIT_LOG_PATH— JSONL audit log pathSLACK_MCP_DRAFT_TTL_SECONDS— draft expiration (default 3600)SLACK_MCP_SCRUB_PROMPT_INJECTION—true/false(defaulttrue)
Safety patterns
- Draft + confirm. Every send (
send_message,send_reply_quote,update_own_message) returns adraft_id. Nothing posts untilconfirm_send(draft_id)is called. Drafts expire after 1 hour. One-time confirm.update_own_messagedrafts include a before/after preview so the operator can diff before confirming the edit. - Workspace required on writes. No global default for sends — every write tool requires
workspaceto prevent wrong-workspace posts when channel names collide. - Audit log. Every tool call appends a JSONL record. Tokens are redacted.
- Prompt-injection scrubber. Incoming message text is sanitized for known prompt-injection patterns (zalgo, role-spoof headers, fake fences). Hits are wrapped not deleted so the operator sees the attempt.
- Token redaction.
list_workspacesreturns redacted profiles only. Tokens never appear in tool responses.
Related MCPs
Same author, same architecture pattern (FastMCP, draft+confirm on writes where applicable, vault auto-export, MIT):
- imessage-mcp - macOS iMessage
- whatsapp-mcp - WhatsApp via whatsmeow
- google-workspace-mcp - Gmail / Calendar / Drive / Docs / Sheets
- apollo-mcp - Apollo.io CRM + sequences
- substack-mcp - Substack writing + analytics
- luma-mcp - lu.ma events
- parse-mcp - markitdown / Docling / LlamaParse router
- rescuetime-mcp - RescueTime productivity data
- graph-query-mcp - vault knowledge graph queries
- investor-relations-mcp - seed-raise pipeline tracker
- vault-sync-mcp - bidirectional vault sync
Telemetry
This plugin sends a single anonymous install signal to myceliumai.co the first time it loads in a Claude Code session on a given machine.
What is sent:
- Plugin name (e.g.
slack-mcp) - Plugin version (e.g.
0.1.0)
What is NOT sent:
- No user identifiers, names, emails, tokens, or API keys
- No file paths, message content, or anything from your work
- No IP address is stored after dedup processing
Why: Helps the maintainer know which plugins people actually install, so attention goes to the ones that get used.
Opt out: Set the environment variable MYCELIUM_NO_PING=1 before launching Claude Code. The hook will skip the network call entirely. Already-pinged installs leave a sentinel at ~/.mycelium/onboarded-<plugin> — delete it if you want to reset state.
License
MIT — see LICENSE.
Source: README.md at commit e9c5a02
Tools
0Version history
1- v0.1.1LatestSep 16, 2026
Score
50% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 days25 pts
Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.
- Auth declared10 pts
Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts