Revnu
By Revnu
Overview
Revnu runs a company's growth: it finds buyers, reaches them in the founder's voice, runs paid ads, and gets the business showing up in AI answers. This app brings that work into ChatGPT. Ask what's waiting and the review queue opens in the chat as cards. Each one is a drafted reply, a blog post ready to publish, or an ad budget change. You approve, send, snooze, request changes or dismiss it without leaving the conversation. Nothing goes out without your say-so, unless you've told Revnu that kind of work may ship on its own. What you can do from ChatGPT: - Review queue: list, open, approve, send, snooze, request changes on or dismiss cards. - Hand over a job: ask Revnu to do something ("find 20 fintech CFOs in London", "draft a post on our pricing change") and check how it's going. - Results: read what each channel produced (meetings, replies, ad spend and conversions, pages ranking). - Pipeline: list leads, open one, set its labels. - Block list: see, add and remove people and domains Revnu must never contact. - Workflows: create, change, pause, resume or delete recurring work, and read its runs. Every tool acts on your own Revnu account, with the permissions you grant when you connect. Each connected app gets its own token, which you can revoke in Settings → API Tokens.
Tools
agent_status
ChatGPTagent_thread
ChatGPTask_revnu
ChatGPTwaitSeconds (default 60, max 200) and returns status (queued, running, replied, failed) and replies. replies can hold progress updates while the status is still running; the work is done only once the status is replied. Work the agent wants approved lands in the review queue (review_list). Give the whole task in one message. While an earlier request from this connection is still queued or running, a new message is not sent: the call waits on the earlier request and returns its progress with notSent. Poll it with agent_message_status. Pass interrupt: true only when the customer asks to change or stop the running work; the agent then drops it and starts on the new message. Pass queue: true to hand over another job without stopping the running one: it is sent now and the agent starts it once the current work is done. Pass a clientRequestId so a retry of the same request is safe.blocklist_add
ChatGPTacme.com, jane@acme.com, Jane <jane@acme.com>, linkedin.com/in/jane-doe) so the agent never contacts them. Returns how many were added, how many were already blocked, and the values it could not parse. Needs leads:write.blocklist_list
ChatGPTnextCursor for the next page. Read-only.blocklist_remove
ChatGPTleads:write.customer_sync
ChatGPTlead_get
ChatGPTcontactsTruncated says when). Read-only. In hosts that support MCP Apps, the user also sees this result in the Revnu app.lead_note_add
ChatGPTleads:write.lead_update
ChatGPTvertical, archetype or tags (tags replace the whole list). Pass null to clear a field; omitted fields are left alone. Changes labels only, never status or outreach. Needs leads:write.leads_list
ChatGPTnextCursor for the next page. Read-only. In hosts that support MCP Apps, the user also sees this result in the Revnu app.linkedin_person
ChatGPTaccount): whether they're connected, any pending invite and when it was sent, Revnu's own invite record with when it saw acceptance, and the newest 50 messages with who sent each. profile is a linkedin.com/in/ URL or id. Read live from LinkedIn: each account is one paced profile lookup, so look people up as you need them rather than in bulk. Read-only.results_get
ChatGPT{available: true, data} or {available: false, reason}. Zero and unknown are different. since is epoch milliseconds, clamped to the last 31 days; default 24 hours. In hosts that support MCP Apps, the user also sees this result in the Revnu app.review_answer
ChatGPTapprove for drafts, budgets and proposals; answer with text for questions; choose with exactly one optionValues entry for choices. Only use a verb listed in the card's actions. Gmail cards return 409 use_send: call review_send instead. This is a real approval: a blog draft publishes, a Reddit post posts, an outbound proposal sets up.review_dismiss
ChatGPTreason (up to 2,000 characters) teaches the agent what to change next time.review_get
ChatGPTreview_list
ChatGPTactions, the exact verbs the API accepts for it, and primaryLabel, what the dashboard's main button would say. One call returns up to 50 open cards (pass limit for up to 100; any beyond that are not returned): pending ones and ones the agent is revising together, with status on each card saying which. There is no status filter. In hosts that support MCP Apps, the user also sees the cards, with their buttons, in the Revnu app. Read-only.review_request_changes
ChatGPTreview_search
ChatGPTreview_send
ChatGPTbody to send an edited version instead of the draft. This sends real email from the customer's mailbox.review_snooze
ChatGPTuntil (epoch milliseconds, in the future). Cancels any auto-ship deadline on it.whoami
ChatGPTworkflow_create
ChatGPTscheduleText is plain words (every weekday at 09:00 Europe/London); each step is a title plus optional details lines, and gate: { on: true } makes that step wait for a human ok. Put volume in the details as words (five per day). The workflow is live on creation; read it back with workflows_list for nextRunAt, or scheduleError when the schedule was not understood. Needs workflows:write.workflow_delete
ChatGPTworkflows:write.workflow_pause
ChatGPTworkflows:write.workflow_resume
ChatGPTworkflows:write.workflow_run
ChatGPTid; it starts within about a minute, and workflow_runs shows its progress. Pass a clientRequestId so a retry is safe. Needs workflows:write.workflow_runs
ChatGPTworkflow_update
ChatGPTsteps, when passed, replaces the whole list. Fields left out are unchanged. Needs workflows:write.workflows_list
ChatGPTCapabilities
Example Prompts
Click any prompt to copy it.






