AI agents and MCP
Ask your AI assistant for a post. Your team still approves it.
Connect ChatGPT, Claude, Grok, Cursor or any MCP client to Wavefront. Agents plan, draft and revise in your workspace with the same documents, checks and rules as the web app. Approval stays with people.
https://wavefront.adwave.com/mcp Connect
Works with the assistant you already use
Add Wavefront as a custom connector in a few steps. Each guide has the exact menus and what to expect.
ChatGPT
Ask ChatGPT to plan, draft and publish approved work in Wavefront, through the MCP connector.
Setup guide CClaude
Add Wavefront as a Claude connector to plan, draft and publish approved work from a chat.
Setup guide GGrok
Connect Grok to Wavefront as a custom connector, or a Grok bot on the xAI API with an API key.
Setup guide CCursor
Add Wavefront to Cursor's MCP servers to draft, revise and export articles from your editor.
Setup guideAny client that speaks MCP over Streamable HTTP works too. See every option.
The safety model
Agents do the work. People make the calls.
Everything an agent does is recorded as its own and checked like any other draft. Actions that reach readers or change who can do what need a person in the web app.
What agents can do
- Talk to the strategist, edit and save plans, and draft calendar posts
- Start articles, copy and images, answer their questions and request revisions
- Read documents, versions, comments, the calendar, search performance and site metrics
- Add backlog ideas and preview the editorial mix
- Comment and ask a teammate to review
- Publish or schedule versions that are already approved
What stays with people
- Approve a version
- Post to social networks or send a newsletter
- Withdraw a publication
- Turn on auto-approve, change billing or manage the team
- Enter destination credentials
Ideas
Things to ask
Agents read your plan and documents first, so short requests work.
- Plan next month for our blog: two posts a week, mostly how-to guides.
- What's waiting for my review in Wavefront, and what's due this week?
- Draft Friday's LinkedIn post about the new cold brew kit in our usual voice.
- Which of our pages are close to page one, and what would you refresh first?
- Add five local-intent keywords for Austin and two posts a week for them.
- Revise the pour-over article: add a grind-size table and shorten the intro.
OAuth for people
Your assistant registers itself and you sign in to Wavefront in a pop-up. Read, write and stay-connected permissions are listed before you choose Allow. The connection lasts 30 days, and you can disconnect it any time under Account → Connected agents.
API keys for bots
An owner creates a key with the scopes it needs in Settings → API and MCP. The bot sends it as a bearer token. A key reaches only its own workspace.
Every tool, documented
Each tool has a title, a description and safety hints, so clients can ask before they act. Browse the tool reference.
Questions
What is MCP?
The Model Context Protocol is an open standard that lets AI assistants call tools in other services. Wavefront runs a remote MCP server, so assistants that support MCP can work in your workspace. See connecting AI agents to your content workflow.
How does an agent sign in?
Most clients use OAuth: you sign in to Wavefront in a pop-up and choose Allow, and your role in the workspace applies. Bots and scripts can use a workspace API key as a bearer token instead, and the key's scopes decide what it can do.
Can an agent publish without anyone checking?
No. An agent can publish only approved versions: approved by a person in the web app, or by autopilot when an owner has turned on auto-approve. It can't post to social networks, send newsletters or withdraw anything.
Does using an agent cost extra?
No. MCP and the API are included in the plan. Drafts an agent starts use credits like any other. See pricing.
Give your assistant a workspace that checks its work
Start a 7-day trial with 300 credits and every feature. No card needed. Then $99 a month.