Your agent works Viably the way you do.
Connect Claude, Cursor or Codex and sign in once. The agent reads your inbox, the pipeline, the board and the relay logs, then claims contacts, moves deals and files tasks after you approve each change. It acts as you, so its work shows up in the dashboard and in Slack under your name.
- 14:01:58Claude.aisearch_console_summary Clicks up 18% on last weekRead
- 14:01:46Codexlist_deals 5 deals stalled in ProposalRead
Add the server, then sign in as yourself.
Each app adds Viably with one command or one config entry, then opens a Viably sign-in in your browser. You pick the workspace, see what the agent may read and change, and approve. The app keeps its own token and renews it.
claude mcp add --transport http viably https://api.viably.app/mcp- 1Add the server
- 2Sign in
- 3Ask
One command adds the server. Run /mcp inside Claude Code, pick viably, and approve the sign-in it opens.
What the agent changes lands where your team already looks.
Each change runs the dashboard's own write as you. A claimed contact shows you as its owner, a filed task posts its card to Slack, and a replayed event reaches its channel again. The audit trail records every one with the app that made it.
/mcp__viably__triage_inboxChanges
API and agentsThe saved triage prompt reads your queue, shows a plan, and asks before each change. The reply stays a draft on the record, because sending it is yours.
Reads start ticked. Changes wait until you tick them.
Each area of the workspace is a scope you grant at sign-in, and an agent that holds only reads cannot change a thing. The consent screen below works: tick a change and watch which tools it opens.
Connect Claude Code to Viably
Untick anything it does not need. Try it.- whoami
- search
- list_inbox
- find_contacts
- get_contact
- list_deals
- list_tasks
- list_issues
- get_issue
- list_pins
- query_relay_logs
- relay_endpoints
- medal_leaderboard
- list_events
- list_reminders
- search_console_summary
- social_summary
- uptime_status
- github_activity
- list_channels
- create_task
- update_task
- create_issue
- resolve_issue
- claim_contact
- set_contact_status
- tag_contact
- add_contact_note
- open_deal
- move_deal
- dismiss_inbox_item
- snooze_inbox_item
- set_reminder
- cancel_reminder
- create_event
- replay_relay_event
- Reply to a customer
- Merge or block contacts
- Change billing
- Change members or powers
- Give medals
- Your role and no more
Every call runs through the same permission checks as the dashboard, as you. Scopes can narrow what the agent reaches.
- A contact list counts as a download
Listing names through find_contacts needs the download permission, and each list is logged under Downloads like any other export.
- Outside text is treated as data
Form fields, emails and webhook payloads reach the model as the agent works. Viably labels them as data in every result, and keeps your app asking before each change.
Saved prompts, and records you hand over yourself
The server ships three prompts that chain the tools the way a member works the dashboard. In Claude Code they are slash commands, like /mcp__viably__triage_inbox, and each one shows you its plan before it changes anything.
Type @viably: to attach a contact waiting on a reply or an open issue. The record loads with its deals and timeline before the agent starts, masked the way your role masks it.
/mcp__viably__Claude reads your queue with list_inbox and proposes an ending for each item. First up: Michael Scott has waited two days on a reply.
Type a slash for the saved prompts, which chain the tools the way you work the dashboard and mention only the changes your sign-in allows. Type @viably: to attach a person or an open issue, so the agent starts with the record in hand.
Agents take turns, so the dashboard stays quick
A workspace runs 2 tool calls at a time. The rest wait for up to 10 seconds, and a call that waits longer answers 503 with a Retry-After header, so an agent firing a burst slows itself down and nobody else.
Tool calls count against the workspace’s monthly allowance: 500 on Free and 15,000 on Viably+, plus 3,000 for each top-up pack. Each token also gets 30 requests a minute on Free and 60 on Viably+, where a change costs 3.
Limits and the audit trailSetup and stories for each app
Every app reaches the same tools. Each page shows what that app is best at, from triaging the inbox in Claude to debugging a webhook in Cursor.
Questions about the MCP server
What is the Viably MCP server?
Which apps can connect?
Can my agent see anything I can't?
Can it email or message our customers?
What if a customer's message tries to instruct my agent?
Does it cost extra?
Ask your agent what is waiting on you.
Start free, open API and agents, and copy the command for your app. Your inbox has an assistant the same afternoon.
- Free to start
- No credit card required
- Powers on when you need them