MCP and Skills
Use Volae from your own agent

Connect Claude Code or Codex to our remote MCP server and Skills, or script the volae CLI. Your local agent runs the same actions as the web app, under the permissions you set.
One command to connect Run it in your terminal, then sign in from the browser.
$ claude mcp add --transport http volae https://api.volae.net/mcp
- 1Run /mcp in Claude Code, choose volae, and sign in from the browser
- 2New connections get read & write access. They can read everything and save drafts or proposals, without permission to approve or send
- 3Volae's tools show up in Claude Code
Remote server Sign in with OAuth, no pasted keys.
Too many tools? Connect the compact server. It exposes only three.
/mcp/compactsearch_toolscall_readcall_writeTwo ways to generate
Let us write it, or just take the context

Every writing tool has two modes. Have Volae write the draft, or pull the context and run it through your own model and subscription.
Volae generates Call draft_reply and our platform models write the draft. You pay for actual usage.
per reply draft
About9credits
draft_replyContext only, your model writes Pull what the reply needs to your machine, write it, and save it back.
- 01
Pull the material
The thread, customer account, product memory and matching knowledge
get_draft_context - 02
Your model writes
Locally, on your own subscription
- 03
Save it back
As a draft or proposal. Sending still needs your approval
draft_reply
Inside Claude Code
Reads and drafts run. Destructive actions come back to you

Checking today, pulling context and saving drafts all happen in the terminal. Anything that goes out gets approved in the web app.
What needs me today?
volae · get_standup
2 need you: ChatGPT still says the bag isn't compostable, and Lin Xia wants to pause.
Draft a reply to Lin Xia. I'll write it, don't use credits
volae · get_draft_context
Got the thread, account, product memory and 3 relevant knowledge entries. Draft written.
volae · draft_reply → saved
Saved as a draft. Sending needs your approval in the web app.
Runs right away Reading data, pulling context, saving drafts and proposals.
Comes back to you Sending a draft or deleting anything waits until you confirm it in the web app.
Skills
The daily work, written down as Skills

The same SKILL.md files work in both Claude Code and Codex. Point your agent at a file and it knows which Volae tools to call to finish the job.
Growth
6geo-analysisChecks how AI assistants answer buyer questions and whether they mention you.
seo-auditBreaks down site audits and search rankings. Orders the fixes by impact.
seo-autopilot-cadenceSets up a weekly routine for SEO and AI search.
competitor-watchTracks rival price changes and changelogs, plus how people talk about them.
content-briefPicks the next topic to write and drafts the brief.
outreach-writingFinds the right contacts and drafts the first email plus follow-ups.
Every day
3daily-briefPulls today's brief and walks through the items waiting on you.
theme-synthesisSummarizes customer themes and the evidence behind them.
volae-knowledgeUpdates the knowledge base directly from your code, config, and docs.
Retention
3faq-from-supportFinds repeated support questions and turns them into knowledge base entries.
warm-emailDrafts personal check-ins and follow-ups for existing customers.
interview-guidePicks users to interview and writes the invite and guide. Pulls quotes from the transcript.
Skills ship with the Claude Code plugin, which is published at launch.
Permissions
Connected doesn't mean anything goes

Your agent signs in with your Volae account. By default it can read and draft. Sending and deleting come back to you in the web app.
Read & write by default, nothing outbound
Allowed once connected
- Read all your data
- Save drafts and proposals
You turn these on in the web app
- Approve
- Send to anyone
Approving and sending are switched on per agent in the web app, and turning them on sends you a security email.
OAuth sign-in, no pasted keysYour agent signs in through your Volae account. Disconnect it and its next call returns 401 right away.
Destructive actions come back to youDestructive calls pause until you confirm them in the web app.
Every call on recordThe audit log and change log record which agent ran, when, what it called, and what changed.
One level down
Your models and data sources, too

Paid plans only.
Bring your own model
Register an endpoint that speaks the Anthropic Messages format, plus its key. It has to pass our checks at registration, then you pick your model or a platform model per task type.
Checked at registration
Safety steps always run on ours
- Deciding when to ask you
- Filtering instructions out of outside content
- Compliance checks before publishing
Custom data sources
Your sources run alongside Volae's built-in data, which stays in place. The coworker and Lookup can call them and map results to evidence or person and company fields.
What you can connect
Everything returned is treated strictly as data, never as instructions.

14-day trial, no card
Plug Volae into your own agent
Create an account, run the command above, and sign in from the browser.
Includes 500 credits. No charge when the trial ends, your account turns read-only and your data is kept for 30 days.





