Integrations - Everflow API

EU Cluster Support: If your Everflow account is hosted in the EU, simply replace https://mcp.eflow.team with https://mcp-eu.eflow.team in any of the configuration tables and steps below.

These integrations let you bring Everflow data into the tools your team already works in — without opening the Everflow portal or writing API calls.

Slack

Claude in Slack is now Claude Tag — Anthropic replaced the earlier Claude in Slack app on 3 August 2026. It’s on Team and Enterprise plans, in public beta. How Everflow reaches it depends on where you’re asking, and the two paths are genuinely different:

Where you ask	Whose connector is used	Who sets it up
Direct message with Claude	Yours, from your own claude.ai account	You. Nothing else needed
Channel or thread (`@Claude …`)	The channel’s, configured by an admin	An Owner in your Claude organization

If you @Claude in a channel and get a notice offering to “set up Claude Tag” instead of an answer, your organization hasn’t run the one-time setup yet. Personal connectors don’t apply in channels — that’s the admin path below.

In a direct message (works today)

Personal connectors keep working in DMs, so if you’ve already connected Everflow to your claude.ai account there’s nothing more to do.

Connect Everflow to your own Claude account

Follow the Claude.ai setup in the Quickstart — Settings → Connectors → Add custom connector, authentication None, and an x-api-key request header holding your Network API key.

DM Claude in Slack

Open a direct message with Claude and ask normally:

what were our top 5 partners by revenue last week?

did offer 88 hit any caps yesterday?

A DM runs on your own seat rather than the organization’s usage balance, and it sees exactly what your API key allows.

In a channel (needs an Owner)

Channel access is granted per channel by an admin, not per person. Claude Tag reaches an MCP server through two pieces: a plugin that tells it the server exists, and a credential that lets the call leave the sandbox with authentication attached. You need both — the credential alone opens the network path, but without the plugin Claude never learns there’s a server there to call.

Searching the tool gallery for “Everflow” returns nothing, and that’s expected. That list is Anthropic’s preset integrations; Everflow is a custom remote MCP server and will never appear in it. Skip the “Choose Claude’s first tools” step in the setup wizard and add Everflow through Custom tool on the Credentials tab instead.

An Owner runs the Claude Tag setup

At claude.ai/admin-settings/claude-tag. Only a Primary Owner or Owner can — the Admin role can’t. The wizard runs five steps: pair the Slack workspace, choose Claude’s first tools, connect GitHub, create accounts for Claude’s other tools, and launch.Pairing is: add the Claude app to Slack, send @Claude connect in any channel, then paste the workspace_… code it replies with back into the setup page.Everflow goes in step 4, “Create accounts for Claude’s other tools.” It isn’t a preset, so there is nothing to pick in step 2 — skip that step.

Add a plugin pointing at the Everflow MCP server

On the Access bundle’s Plugins tab, add a plugin whose .mcp.json points at https://mcp.eflow.team (or https://mcp-eu.eflow.team). An .mcp.json committed to a repository is not picked up — it has to arrive through an attached plugin.

Add the credential for the Everflow host

On the same bundle’s Credentials tab, choose Connect → Custom tool with these values:

Field	Value
Credential type	Bearer — the only type that exposes the Custom headers field
Custom headers	`X-Api-Key: <your-network-api-key>`
Allowed websites	`mcp.eflow.team` (or `mcp-eu.eflow.team`)

Everflow authenticates on the X-Api-Key header, not on Authorization: Bearer, which is why the credential goes in Custom headers.

Scope the bundle to the right channels

Attach the bundle only to the channels that should see network data. A bundle attached at the workspace or organization level reaches every channel under it.

Verify in a new thread

@Claude what can you access from this channel?

@Claude pull yesterday’s top 5 partners by revenue from Everflow

The credential is Claude’s account, not yours — anyone in a channel under that bundle can use it, and channels are the widest audience an Everflow key gets. Create a dedicated Network API key for the agent, Read Only on just the modules those channels should see. Think twice before granting Reporting to a broad channel: it can return consumer emails, IPs, and mobile ad IDs. See Choosing a key for a connector.

Anthropic’s custom connection guide covers the console mechanics; the table above is the Everflow-specific part.

Scheduled tasks

Once the connector is on your Claude account, Claude’s scheduled tasks can run a prompt against it on a repeating schedule — hourly, daily, weekly, or weekdays only — and deliver the result somewhere. That turns a read-only query interface into recurring reporting without a line of code, a cron host, or a workflow tool. The scheduling lives in Claude, not in Everflow. Everflow MCP has no scheduler of its own and never pushes data — see the note on event systems.

Connect Everflow, plus wherever the output should land

The Everflow connector supplies the data. Delivery needs its own connector — Gmail or Outlook to email it, Slack to post it into a channel. A task can also just write the result to a file or leave it in the Claude app.

Set the tools the task needs to run without asking

This is the step people miss. Any tool left on Needs approval will stall a scheduled run, because nobody is there to approve it. In Settings → Connectors → Everflow MCP → Tool permissions, set the tools the task actually uses to allow — for a performance digest that’s run_network_summary, run_performance_report, and usually get_account_info.

Describe the job and pick a cadence

Ask Claude for the task in plain language and it will set up the schedule:

“Every weekday at 8am, pull yesterday’s network summary from Everflow with a comparison against the previous day, list the five offers with the biggest revenue swing, and email me the summary.”

Claude saves the prompt as the task’s instructions and re-runs it on the cadence you choose.

Check the first run

Read the first result before trusting the schedule. Confirm the numbers match the portal for the same window — the reporting tools return an applied_query object showing the exact resolved query, including the timezone and currency actually used.

Jobs worth scheduling

Cadence	Task
Hourly	Cap and pacing watch — `list_entities(type="offer_cap")` for offers past a consumption threshold, reporting only when something is near or over
Daily	Performance digest — yesterday vs the day before, top movers by `revenue_pct_change`, with a note on what changed
Daily	Traffic health check — `list_traffic_health(type="tasks")` for open remediation work, and certificates expiring in the next 30 days
Weekly	Partner review — new signups from `list_affiliates(status="pending")`, plus partners whose EPC moved most week over week
Weekly	Traffic quality — invalid-click rate by partner and sub-source, ranked by rate rather than volume

Scheduled runs are a good fit for Agent Skills. A task that says “run the performance-review skill for yesterday and email it” gets the skill’s full procedure and interpretation on every run, instead of you re-specifying the analysis in the prompt each time.

What a bad schedule can and can’t do. The MCP Server is read-only, so a scheduled task cannot create, update, or delete anything in your network however the prompt is worded — the worst outcome for your data is a wrong summary. Disclosure is the risk that’s left: the task reads whatever its key allows and delivers it wherever you pointed it. Give it a dedicated key scoped to just what the digest needs, and check where the output lands before you leave it running. See Choosing a key for a connector.

Once Claude Tag is set up, it can also run scheduled jobs and watch channels from inside Slack, so a recurring digest can post to a channel rather than land in your inbox. Task availability, cadences, and the exact setup screens differ by Claude plan and surface. See Anthropic’s guides to scheduling recurring tasks and Claude Tag for current details.

Zapier

Use Zapier to build automated workflows that query Everflow via AI and post results to Slack, email, Google Sheets, or anywhere else. This works with both the Claude AI action in Zapier and Zapier’s native MCP support.

Daily performance summary to Slack

A common workflow: schedule a daily summary of the previous day’s performance and post it to your team’s Slack channel.

Create a new Zap with a Schedule trigger

Set the trigger to Schedule by Zapier, configured to run every day at your preferred time (e.g. 9:00 AM).

Add a Claude AI action

Add an Anthropic Claude action and configure the prompt:

You are connected to the Everflow MCP server. Query yesterday's performance
using run_performance_report with dimensions=affiliate, date range = yesterday.
Return a concise summary of the top 5 affiliates by revenue, including
clicks, conversions, and payout. Format it for a Slack message.

Add a Slack action to post the summary

Add a Send Channel Message action in Slack, using Claude’s response as the message body. Choose the channel where your team reviews daily numbers.

Fraud alert workflow

Trigger an AI investigation when a conversion comes in above a certain payout threshold.

Set a webhook trigger

Use Webhooks by Zapier as the trigger. Configure your Everflow network to fire a postback to the Zap’s webhook URL on high-value conversions.

Add a Claude AI action to investigate

Pass the conversion ID from the webhook payload to Claude:

Using the Everflow MCP server, investigate conversion {{conversion_id}}.
Pull the full conversion details, trace the originating click, and check
whether there are duplicate conversions from the same affiliate and order ID.
Flag anything suspicious.

Post the finding to Slack

Route Claude’s analysis to a dedicated fraud-review Slack channel so your team can act immediately.

n8n

n8n is a self-hosted workflow automation tool popular in the performance marketing and agency space. Its native MCP Client node connects directly to the Everflow MCP server, making it straightforward to build automated reporting and alerting pipelines.

Add an MCP Client node to your workflow

In your n8n canvas, add an MCP Client node. Configure it with the Everflow server details:

Field	Value
Server URL	`https://mcp.eflow.team`
Authentication	Header
Header name	`X-Eflow-API-Key` (or `X-Api-Key`)
Header value	`your-network-api-key`

Choose a tool and pass parameters

Select the Everflow tool to call (e.g. run_performance_report) and configure the input parameters. n8n will discover available tools automatically from the MCP server.

Connect downstream nodes

Route the MCP response to any n8n node — Slack, Google Sheets, email, a database, or an AI node for further processing. A common pattern for agencies is: MCP → AI summarize → Google Sheets → Slack notification.

Use n8n’s Schedule Trigger node to run Everflow queries on a recurring cadence — daily, weekly, or at the end of each billing period — and push results to your reporting stack automatically.

Raycast

Raycast is a launcher and productivity tool popular with operators and power users. Its built-in AI supports MCP, letting you query Everflow from your desktop without switching to a browser or opening the portal.

Open Raycast AI Settings

Launch Raycast and go to Settings → AI → MCP Servers.

Add the Everflow server

Click Add Server and enter:

Field	Value
Name	`Everflow`
URL	`https://mcp.eflow.team`
Header	`X-Eflow-API-Key: your-network-api-key` (or `X-Api-Key`)

Query Everflow from the Raycast bar

Open Raycast (⌘Space), launch AI Chat, and ask:

What were yesterday’s top offers by conversion rate?

Is affiliate 99 currently active and when did they last convert?

Microsoft Copilot Studio

For teams on the Microsoft 365 stack, Copilot Studio lets you build a custom Copilot agent that surfaces Everflow data inside Teams, Outlook, and other Microsoft surfaces. See Microsoft’s setup guide for the full walkthrough on adding an MCP server to an agent.

Create a new agent in Copilot Studio

Go to copilotstudio.microsoft.com and create a new agent. Give it a name like “Everflow Assistant.”

Add the Everflow MCP server as a tool

Under Actions → Add an action → Model Context Protocol, add:

Field	Value
Server URL	`https://mcp.eflow.team`
Header	`X-Eflow-API-Key: your-network-api-key` (or `X-Api-Key`)

Publish and connect to Teams

Publish the agent and add it to your Microsoft Teams environment. Team members can then @mention it in Teams channels to query Everflow data.

Microsoft Copilot Studio requires a Power Platform license. Check with your Microsoft administrator if you’re unsure whether your organization has access.

​Slack

​In a direct message (works today)

​In a channel (needs an Owner)

​Scheduled tasks

​Zapier

​Daily performance summary to Slack

​Fraud alert workflow

​n8n

​Raycast

​Microsoft Copilot Studio

Slack

In a direct message (works today)

In a channel (needs an Owner)

Scheduled tasks

Zapier

Daily performance summary to Slack

Fraud alert workflow

n8n

Raycast

Microsoft Copilot Studio