Slack MCP server and Dome
Let agents read Slack, and decide where they post. Connect Slack's MCP server to Dome once. Every call runs with the Slack user token of the person the agent acts for, and rules decide each message on its channel.
Dome provides a tool gatewayand authorization for Slack.
Decided on the channel
The same tool is allowed in one channel and refused in another, because rules read the channel_id of every message.
Their token, their channels
Each teammate authorizes your Slack app once, on their first call. An agent acting for them reads only what they can.
Reads stay open
Search and read calls pass while posts are narrowed. A prompt can't send an agent's message to #general.
Slack behind the Gatewayin three steps.
Slack's MCP server takes an internal or directory-published Slack app, with no dynamic client registration. Add the server with the app's client ID and secret, sync, and apply the rules.
Add the Slack MCP server
Turn on MCP for your Slack app and give it the user token scopes your agents need. Credentials are per user, through OAuth.
shelldome tools add --name slack \ --url https://mcp.slack.com/mcp \ --auth-method oauth --credential-type per-user \ --oauth-client-origin manual \ --oauth-client-id "$SLACK_CLIENT_ID" \ --oauth-client-secret "$SLACK_CLIENT_SECRET" \ --oauth-authorize-url https://slack.com/oauth/v2_user/authorize \ --oauth-token-url https://slack.com/api/oauth.v2.user.access \ --oauth-token-endpoint-auth client_secret_post \ --oauth-default-scope search:read.public \ --oauth-default-scope channels:history \ --oauth-default-scope chat:write \ --gateway ops-gatewaySync the catalog
Sync spends your own Slack credential, so authorize the app first. Until then it fails and agents get “tool not available in this gateway”.
shelldome tools catalog sync slackApply the rules
Scope them to one agent while you try them, and simulate before you deploy.
shelldome rules apply slack-args.cedar \ --agent escalation-bot --name slack-per-channel
Read anywhere,post in one channel.
The permit opens Slack to the agent. The forbid refuses sent and scheduled messages to any channel but #support-escalations.
cedar
permit (principal, action == Dome::Action::"mcp:call", resource is Dome::MCPTool)when { resource.connection_name == "slack" };
forbid (principal, action == Dome::Action::"mcp:call", resource is Dome::MCPTool)when { resource.connection_name == "slack" && ["slack_send_message", "slack_schedule_message"].contains(resource.tool_name) && resource has arguments && resource.arguments has channel_id && resource.arguments.channel_id != "C04ESCALATE"};One call,two outcomes.
Switch the caller or the argument and watch the same call decide differently, with every decision landing in audit.
Channel
agent escalation-bot · acting as l.chen
slack/slack_send_message(channel_id: "C04ESCALATE", message: "Ticket 4821 breached its SLA")
- Agentescalation-bot is registered and active
- Callerl.chen verified through Okta
- RulePosts to #support-escalations are allowed
DecisionAllowed
Three agentson Slack.
These come from our template library, and each one reads broadly and writes narrowly.
thread-summarizer
Summarize long threads
Reads a long thread and drafts a summary for the person who asked, who then decides whether to send it.
escalation-bot
Escalate breached tickets
Posts tickets that breach their SLA to #support-escalations, with the customer's recent history.
standup-digest
Schedule a standup digest
Reads yesterday's team channels and schedules a morning digest in the team's own channel.
Bringing it alltogether.
Connecting Slack to registered agents, models, and identity in Dome completes a governed agent application.
Acting for
Control point
GatewayEvery call decided and audited
RulesGuardsQuotas
Tools
SlackMCP server
Models
Common questionsabout Slack.
More of whatDome works with.
Model
Claude Fable
Fable 5.1 from Anthropic and Amazon Bedrock in one failover pool, open to one group and capped by quota.
Provider
Anthropic
The Claude API behind the Model Broker: the key held in Dome, every call authorized, metered and audited.
Identity
Microsoft Entra ID
Entra access tokens verified on every agent call, so rules read app roles and audit names the person.
Runtime
LangGraph
LangGraph agents with their model calls on the Model Broker and their MCP tools on the Tool Gateway.
Client
Claude Code
Claude Code on a Dome Gateway with per-developer sign-in, rules on every tool call, and audit by name.
Agent service
TinyFish
TinyFish's web agents behind the Tool Gateway, with rules that decide each run on the site it targets.
Talk with ourFDE team.
Or start free, and register your first agent in minutes. Connect it through a gateway, give it rules, and see every call it makes in one audit trail.
Talk with our FDE team
Our engineers design, build and run governed agents for the part of your business you want to change. You see what they deliver.
MCP servers for AI agents: per-argument rules on every call
Put any remote MCP server behind the Tool Gateway. Agents reach it through one governed endpoint, and rules decide each call on its arguments.