MCP over OAuth
One hosted endpoint at api.oviond.com/mcp on streamable HTTP transport, with OAuth 2.1 and PKCE. Your assistant opens a browser, you sign in to Oviond, and it receives a scoped access token.
Feature
Reporting API and MCP server
Oviond exposes its reporting surface twice: a REST API for your own code, and a hosted MCP server so AI assistants like Claude and ChatGPT can build and manage client reports directly.
Why the API and MCP server
Client reporting, where your team already works
Reporting that happens in a chat window or a script is reporting nobody has to block out an afternoon for.
Reporting inside your AI
Connect Claude, ChatGPT, Cursor, or any MCP client and ask for a client, a report, or a scheduled send in plain language.
The whole surface, twice
Clients, reports, pages, widgets, data sources, templates, themes, automations, and branding are reachable as MCP tools and as REST endpoints.
Scoped and auditable
MCP signs in with OAuth and every call is scoped to that account. REST keys are issued and revoked from Settings, and every response carries a request ID.
How it works
Two ways in
The MCP server is for agents. It speaks the Model Context Protocol over streamable HTTP at api.oviond.com/mcp, authenticates with OAuth, and presents Oviond as tools an assistant can call. The REST API is for your own code, authenticated with a Bearer key.
Both sit on the same data and the same permissions: an agent or a script can only touch what the signed-in user could touch in the app.
The clients that connect
Claude on web and desktop, Claude Code, ChatGPT, Codex, Cursor, VS Code, Windsurf, and any other MCP-compatible client. Setup differs only in where you paste the URL.
What an agent can do
180+ tools spanning clients and folders, reports and pages, every widget type, data sources and connections, templates and themes, automations, branding, and account-wide search.
A skill, not just a tool list
A tool description says what one call does, never what order to make the calls in. Install the Oviond skill and your agent gets the workflows — including which steps it should hand back to a person.
REST API
150+ endpoints over HTTPS at api.oviond.com, a Bearer key from Settings → API Keys, a published OpenAPI spec, request IDs on every response, and a 300-requests-per-minute account limit that never throttles the web app.
A read-only sandbox
Make real calls against a shared demo account with Google Ads, Google Analytics 4, and Facebook Ads connected, before you wire in a key of your own.
How an agent builds a working widget
Field names are discovered, not guessed. That is the difference between an agent that builds a report you can send and one that invents a metric.
✓
Every call is scoped to the signed-in account
✓
Read-only and destructive tools are labeled before they run
✓
Data is fetched server-side right after a widget is written
✓
Automations are created paused, tested, then activated
Oviond API + MCP
AI reporting workflow
1
List the data sources
The agent looks up the identifier for the platform you named.
2
Describe it
One call returns that source's data views, metrics, dimensions, and any settings it can't run without.
3
Add the widget
The widget is written with real field ids, and its series is fetched server-side straight after the write.
4
Refresh on demand
Re-run the query for a new period without rebuilding anything.
What still belongs in the app
Three things the API and MCP server deliberately hand back to a person.
Connecting Google, Meta, LinkedIn, or TikTok
Sign in once in the app; linking and reporting then work over MCP.
Uploading media or a custom dataset
Files go in through the app; MCP lists, renames, and removes them.
Sending real email or deleting records
Create it paused, send a test, and confirm before it goes live.
What an agent is actually building
The API is a way in. These are the three features it drives.
Finish the layout by hand
An agent can create pages and widgets in seconds. Open the report in the builder afterwards to size, title, and annotate them the way the client reads best.
One field catalog, two front doors
The metrics and dimensions an agent discovers over MCP are the same ones the widget builder shows you, so a report built by prompt and one built by hand are identical underneath.
Schedule a report from a prompt
Ask for a monthly send to a client's recipients and the agent drafts the automation, sends a test, and leaves it paused until you confirm.
The sources an agent can report on
The API and MCP server reach the same 70+ marketing data sources the app does, and describe each one's fields on request.
Clone a template from a prompt
Ask for a Google Ads dashboard and the agent finds the matching template, creates the client, and clones the layout into their first report.
Digital marketing dashboard
Track conversions, revenue, traffic, and customer engagement across your client's full marketing stack in one branded dashboard.
Google Analytics 4Google AdsMeta Ads (Facebook Ads)+4
Google Analytics 4
Track sessions, users, and traffic sources with insights by country, device, age, gender, and channel performance analysis
Google Analytics 4
E-commerce
Analyze e-commerce sales, conversions, and traffic insights with interactive, real-time performance metrics.
ShopifyWooCommerceStripe+1
Website performance
Monitor website traffic, user demographics, and page performance with insights from Google Analytics 4, Search Console, and Page Speed Insights.
Google Analytics 4Google Search ConsoleGoogle PageSpeed Insights+2
PPC dashboard
Track and compare reach, clicks, impressions, and CTR across Google, Facebook, Instagram, LinkedIn, and Twitter Ads over time.
Google AdsMeta Ads (Facebook Ads)Microsoft Ads+7
Social media ads
Track impressions, clicks, CTR, and ad spend across Facebook, Instagram, LinkedIn, TikTok, and Twitter Ads with insights on engagement & conversions.
Meta Ads (Facebook Ads)Instagram AdsLinkedIn Ads+3
API and MCP server FAQs
Connecting an assistant, what it can do, keys, limits, and what to try first.
Does Oviond have an MCP server?
Yes. Oviond hosts a Model Context Protocol server at api.oviond.com/mcp on streamable HTTP transport, authenticated with OAuth 2.1 and PKCE. It is hosted by Oviond, so there is nothing to run locally.
How do I connect Claude or ChatGPT to Oviond?
In Claude, open Settings → Connectors, add a custom connector pointing at api.oviond.com/mcp, and sign in when the browser window opens. ChatGPT, Cursor, VS Code, Windsurf, and Codex follow the same pattern — register a remote MCP connector at that URL and complete the sign-in. In Claude Code, run claude mcp add --transport http oviond https://api.oviond.com/mcp, then authenticate with /mcp.
What can an AI agent actually do in Oviond?
180+ tools cover most of the product: creating and organizing clients, building reports and dashboards, adding and rearranging pages, adding every widget type, connecting and linking data sources, discovering a source's metrics and dimensions, applying templates and themes, scheduling and testing automations, and updating branding.
Is API and MCP access included in my plan?
Yes. Oviond MCP server and API access are included on every plan.
Does the REST API cover everything the app does?
Close to it. 150+ endpoints mirror the features in the interface, and the OpenAPI spec documents every one. The exceptions are the same as for MCP: data sources that sign in on the vendor's own website, and file uploads.
How do I authenticate with the REST API?
Generate a key under Settings → API Keys and send it as a Bearer token in the Authorization header. Requests and responses are JSON, fields are snake_case, timestamps are ISO 8601, and every response carries a request ID you can quote when reporting a problem.
Are there rate limits?
Yes. Programmatic API and MCP traffic is limited per account — 300 requests per minute per surface by default — and a throttled request returns a 429 with a Retry-After header. The web app itself is never rate limited.
Can I try the API before I sign up?
Yes. A shared read-only sandbox account with Google Ads, Google Analytics 4, and Facebook Ads connected lets you list clients and query real data. Write requests return a 403, so you need a key from your own account to create anything.
Can an agent connect a data source for me?
Sources that authenticate with an API key, yes. Sources that sign in on the vendor's website — Google, Meta, LinkedIn, TikTok — need a browser, so connect those once in the app. After that an agent can link them to clients, build widgets on them, and refresh their data.
Where to go next
The API builds the report; the rest of the platform decides what it looks like. See marketing reporting and marketing dashboards for the two formats, white-label reporting for branding it, and client report sharing for getting it to the client. Full technical reference: the Oviond developer docs.
Give your AI access to your reporting
Connect an assistant over MCP, or call the REST API from your own code, and let the busywork happen where your team already works.