Getting started

Context7

Up-to-date, version-specific documentation for thousands of libraries and frameworks — your agent stops guessing at APIs.

1. Sign in at Context7 (free tier available) and open the dashboard. context7.com/dashboard
2. Create an API key (ctx7sk-…) and copy it.
3. Paste it here and Connect. Your agent talks to Context7's hosted server — ask it about a library's current API to test.

Exa

Neural web search built for AI — meaning-based results instead of keyword matching. A third web-search option next to Tavily and Brave.

1. Sign up at Exa — new accounts include free credits, no card. exa.ai
2. Open the dashboard's API Keys page and copy a key. (Heads-up: this page sometimes rate-limits automated checks — it loads fine in a normal browser.) dashboard.exa.ai/api-keys
3. Paste it here and Connect. Web search switches on automatically.

Firecrawl

Web scraping and crawling built for AI agents — your agent reads any website as clean text, even JavaScript-heavy ones.

1. Create a free Firecrawl account — the free plan includes one-time starter credits, no card. firecrawl.dev → Sign up
2. In the dashboard, open API Keys and copy your key (fc-…). firecrawl.dev/app/api-keys
3. Paste it here and Connect. Your agent gets the official Firecrawl tools (scrape, crawl, map, extract) plus live web search — ask it to read a specific page to test. Firecrawl Docs: MCP server

Security note: The key only spends your Firecrawl scraping credits — it can't touch anything else in your account. Rotate it any time from the same dashboard page.

Jina Reader

Clean article extraction and web search grounding for AI, via Jina's official hosted MCP server.

1. Get a free Jina API key from the Jina home page — new keys include free credits, no card. jina.ai → API key
2. Copy the key (jina_…), paste it here and Connect. Your agent talks to Jina's official hosted MCP server (mcp.jina.ai) with it — ask it to read a specific page to test. Jina Docs: official MCP server

Security note: The key only spends your Jina reader/search credits — it can't touch anything else in your account. Rotate it any time from the Jina dashboard.

Perplexity

Answer-engine web search and research with citations, via Perplexity's official Sonar MCP server.

1. Sign in at Perplexity and open the API settings. Sonar API access is pay-as-you-go — add a little credit to start. perplexity.ai → Settings → API
2. Generate an API key (pplx-…) and copy it.
3. Paste it here and Connect. Your agent gets Perplexity's official Sonar tools — ask it a current question to test. Perplexity Docs: MCP server

Security note: The key only spends your Perplexity API credits. Rotate it any time from the same settings page.

Web search — Brave

Web search on the independent Brave index. Alternative to Tavily — connect either (or both; Tavily wins ties).

1. Create a Brave Search API account. Heads-up: Brave asks for a card even on the Free plan (2,000 queries/month) — if you'd rather not, use Tavily above. api-dashboard.search.brave.com → Register
2. Subscribe to a plan (Free is fine), then open API Keys and create one. api-dashboard.search.brave.com → API Keys
3. Paste it here and Connect.

Web search — Tavily

Live web search built for AI agents. The easiest way to get your agent searching — free tier, no card.

1. Sign up (or sign in) at Tavily — the free plan includes 1,000 API credits a month, no credit card. app.tavily.com
2. Your API key (tvly-…) is right on the home dashboard — copy it. app.tavily.com/home → API keys
3. Paste it here and Connect. Web search switches on automatically — ask your agent something current to test it.

Wolfram Alpha

Computational answers — math, units, dates, science and curated data — from the Wolfram|Alpha engine.

1. Sign in at the Wolfram|Alpha Developer Portal and click Get an AppID (the free tier covers a couple thousand queries a month). developer.wolframalpha.com → Get an AppID
2. When creating the AppID, choose the LLM API so the AppID is enabled for the API your agent uses. Copy the AppID and paste it here, then Connect. Wolfram Docs: LLM API

Security note: The AppID only spends your Wolfram|Alpha API quota — nothing else. Regenerate it any time from the Developer Portal.

Cal.com

Scheduling data and bookings from your Cal.com account — reachable through the app gateway.

Calendly

Your agent checks your booking links and scheduled events — reachable through the app gateway.

Fastmail

Your agent reads, searches and sends email from a Fastmail mailbox — same toolset as the Gmail integration.

1. In Fastmail, open Settings → Privacy & Security → Integrations → App Passwords (or go straight to the app passwords page) and click New App Password. app.fastmail.com → Settings → App Passwords
2. Give it a name like “OSP agent” and grant it access to Mail (IMAP/SMTP). Click Generate and copy the password Fastmail shows once. Fastmail Help: app passwords
3. Paste your Fastmail address and the app password here, then Connect. Your agent connects over Fastmail's standard mail servers:
   IMAP imap.fastmail.com:993 · SMTP smtp.fastmail.com:465

Security note: An app password grants access to that mailbox over IMAP/SMTP only — not your full account. Revoke it any time from the same Fastmail app-passwords page; your account password never leaves Fastmail. (This runs the same first-party mail server as the Gmail integration, just pointed at Fastmail's hosts.)

Gmail

Your agent reads, searches and sends email from a Gmail mailbox.

1. Turn on 2-Step Verification for your Google account (app passwords require it). myaccount.google.com → 2-Step Verification
2. Open the App passwords page, type a name like “OSP agent”, and click Create. Google shows a 16-letter password once — copy it (the spaces don't matter). myaccount.google.com/apppasswords
3. If that page says app passwords aren't available: 2-Step Verification is off, or (on a work account) your admin has disabled them — Google's help page covers both. Google Help: Sign in with app passwords
4. Paste your Gmail address and the app password here, then Connect. Nothing to enable in Gmail itself — IMAP access is on by default. Google Help: IMAP and Gmail

Security note: An app password grants full access to that mailbox. Consider a dedicated Gmail account for your agent. You can revoke the app password any time from the same Google page — your account password never leaves Google.

Google Calendar

Your agent reads availability and creates, updates and deletes events.

Microsoft 365 / Outlook

Outlook mail and calendar, Contacts, To Do, OneDrive/SharePoint and Teams — one-click connect, no keys to copy.

Coda

Your agent works with your Coda docs, pages and tables.

1. In Coda, open Account Settings and under API Settings click Generate API token. Copy it (shown once). coda.io → Account Settings → API
2. Paste the API key here and Connect. Your agent talks to the Coda MCP server with it. Coda MCP server

Security note: The token acts as you across the docs you can see. Coda also lets you scope a token to specific docs — do that for least privilege. Revoke it any time from the same settings page.

Confluence

Your agent reads and searches your Confluence Cloud spaces — read-only — via the official-grade Atlassian MCP server.

1. Create an Atlassian API token (works for both Jira and Confluence): Atlassian account → Security → API tokens → Create API token. Copy it (shown once). id.atlassian.com → Security → API tokens
2. Your Confluence URL is your site with /wiki on the end — e.g. https://your-site.atlassian.net/wiki.
3. Paste your URL, your Atlassian account email and the API token here, then Connect. Your agent runs the Atlassian MCP server in READ-ONLY mode (no edits). Atlassian MCP server (read-only)

Security note: Defense in depth: the agent always runs the Atlassian server with READ_ONLY_MODE on, so it can search and read but never edit pages. The API token still acts as your account on the spaces it can see — revoke it any time from the same Atlassian page.

Google Docs

Your agent creates, reads and edits Google Docs.

Google Sheets

Your agent creates spreadsheets and reads/writes their data.

Notion

Your agent reads and updates the Notion pages you share with it.

1. Open Notion's integrations page and click New integration. Pick your workspace; type Internal. notion.so → Settings → Integrations
2. On the Configuration tab, copy the Internal Integration Secret (starts with ntn_). Under Capabilities choose Read content — add Insert/Update only if you want the agent writing.
3. Crucial: grant the integration access to pages. On the integration's Access tab pick the pages/databases — or on any Notion page: ••• menu → Connections → your integration. Without this it sees nothing. Notion Docs: create an integration (with screenshots)
4. Paste the secret here and Connect.

Security note: The integration only ever sees pages you explicitly connect it to — Notion's permission model is opt-in per page tree.

Outline

Your agent searches, reads and drafts documents in your Outline (team wiki) workspace.

1. In Outline, open your profile → Settings → API Keys → New API Key. Name it and copy the token (shown once). Outline → Settings → API Keys
2. Paste your workspace URL and the API key here, then Connect. Your agent talks to the Outline REST API with it. Outline Docs: API

Security note: The API key acts as you across the documents your account can see. Revoke it any time from the same API Keys page.

Plaud

Your agent can read your Plaud recordings, transcripts and AI summaries.

Bitbucket

Repos, pull requests and pipelines on Bitbucket Cloud — reachable through the app gateway.

CircleCI

Your agent reads build status and failure logs via CircleCI's official MCP server.

1. In CircleCI, open User Settings → Personal API Tokens and create a new token. Copy it (shown once). app.circleci.com → User Settings → Personal API Tokens
2. Paste it here and Connect. Your agent talks to CircleCI's official MCP server with it. CircleCI Docs: MCP server

Security note: A personal API token acts as you on CircleCI. Revoke it any time from the same settings page.

Docker Hub

Your agent searches Docker Hub and inspects repositories, tags and digests.

1. In Docker Hub, open Account Settings → Personal access tokens → Generate new token. Read-only is plenty for lookups. Copy it (dckr_pat_…, shown once). app.docker.com → Personal access tokens
2. Paste the token here (and your username only if you want the agent to see your private repos), then Connect. Public image/tag lookups work either way. Docker Docs: access tokens

Security note: A read-only PAT lets the agent inspect repos and tags but never push or change anything. The username is only used to reach private repos — omit it for public-only use. Revoke the token any time from the same Docker Hub page.

GitHub

Your agent works with your repos, issues and pull requests via GitHub's official MCP server.

1. Create a fine-grained personal access token (recommended — you choose exactly which repositories it can see). github.com → Settings → Fine-grained tokens → New
2. Pick the repositories, then under Repository permissions grant at minimum:
   Contents: Read-only · Metadata: Read-only (automatic)
3. Optionally add, if you want the agent to manage them:
   Issues: Read and write · Pull requests: Read and write
4. Alternatively, a classic token also works — scope `repo` (private repos) or `public_repo` (public only). github.com → Settings → Tokens (classic) → New
5. Click Generate, copy the token (shown once), paste it here and Connect. Your agent talks to GitHub's official hosted MCP server with it; the tools it gets automatically match what the token allows. GitHub Docs: the GitHub MCP server

Security note: Give the token only the repos and permissions you want the agent to have — a read-only token is plenty for code questions. Revoke it any time from the same GitHub settings page.

GitLab

Your agent works with your gitlab.com repos, issues and merge requests.

1. In GitLab, go to your avatar → Edit profile → Access tokens and add a new token. (This page needs you logged in — it blocks automated checks, but it's GitLab's canonical token page.) gitlab.com → User settings → Access tokens
2. Pick scopes. read_api is plenty for questions; use api if the agent should create issues/MRs:
   read_api (read-only) · api (read and write)
3. GitLab's docs cover the steps with screenshots if the page moved. GitLab Docs: personal access tokens
4. Copy the token (shown once), paste it here and Connect. Heads-up: gitlab.com only for now — tell support if you need a self-hosted GitLab instance.

Security note: Set an expiry date on the token and prefer read_api unless you want the agent writing. Revoke it any time from the same page.

Sentry

Your agent investigates errors and performance issues straight from your Sentry projects.

1. In Sentry, open Settings → Account → API → Auth Tokens and create a new token. sentry.io → Auth Tokens
2. Grant at minimum these scopes (add event:write only if you want the agent resolving/assigning issues):
   org:read · project:read · event:read
3. Copy the token, paste it here and Connect. Sentry's docs cover the agent toolset. Sentry Docs: Sentry MCP

Security note: Stack traces can contain user data from your apps — scope the token to the orgs/projects you're comfortable exposing to the agent. Revoke any time from the same page.

Asana

Your agent reads tasks and projects from your Asana workspace — read-only.

1. In Asana, open My Settings → Apps → Manage Developer Apps → Personal access tokens, and create a new token. Copy it (shown once). app.asana.com → My Settings → Apps → Developer
2. Paste the token here and Connect. Your agent runs the Asana MCP server in READ-ONLY mode (no task edits). Asana MCP server (read-only)

Security note: Defense in depth: the agent always runs the Asana server with READ_ONLY_MODE on, so it can read but never create, update or delete tasks. The token still acts as your account on the workspaces it can see — revoke it any time from the same Asana page.

ClickUp

Tasks, lists and docs from your ClickUp workspace — reachable through the app gateway.

Jira

Your agent reads issues, sprints and boards from your Jira Cloud projects — read-only — via the official-grade Atlassian MCP server.

1. Create an Atlassian API token (works for both Jira and Confluence): Atlassian account → Security → API tokens → Create API token. Copy it (shown once). id.atlassian.com → Security → API tokens
2. Your Jira URL is your site root — e.g. https://your-site.atlassian.net.
3. Paste your URL, your Atlassian account email and the API token here, then Connect. Your agent runs the Atlassian MCP server in READ-ONLY mode (no issue edits). Atlassian MCP server (read-only)

Security note: Defense in depth: the agent always runs the Atlassian server with READ_ONLY_MODE on, so it can search and read but never create or transition issues. The API token still acts as your account on the projects it can see — revoke it any time from the same Atlassian page.

Linear

Your agent works your Linear workspace — issues, projects, cycles — via Linear's official hosted MCP server.

1. In Linear, open Settings → Account → Security & Access and create a new personal API key. You can restrict it — read-only and per-team scoping are both offered at creation. linear.app → Settings → Security & Access
2. Copy the key (lin_api_…, shown once), paste it here and Connect. Your agent talks to Linear's official hosted MCP server with it. Linear Docs: MCP server

Security note: A restricted, read-only key is plenty if you only want the agent answering questions about your issues. Revoke it any time from the same settings page.

monday.com

Your agent works your monday.com boards and items via monday's official API MCP server.

1. In monday.com, click your avatar → Developers → My Access Tokens (or Administration → Connections → API), and copy your personal API token. monday.com Docs: authentication / API token
2. Paste it here and Connect. Your agent talks to monday's official API MCP server with it. monday.com Docs: MCP server

Security note: The API token acts as you on monday.com — it sees what your account sees. Regenerate it any time from the same Developers page.

Todoist

Your agent manages your Todoist tasks and projects via Todoist's official MCP server.

1. In Todoist, open Settings → Integrations → Developer and copy your API token. todoist.com → Settings → Integrations → Developer
2. Paste the token here and Connect. Your agent runs Todoist's official MCP server with it. Todoist MCP server (official)

Security note: The token acts as you across your Todoist account. Reset it any time from the same Developer settings page.

Trello

Your agent reads and manages your Trello boards, lists and cards.

1. Open Trello's Power-Up admin page and create a Power-Up (any name, your workspace) — this is just how Trello issues API keys; nothing gets installed on your boards. trello.com/power-ups/admin
2. On the Power-Up's API Key tab, click Generate a new API key and copy it.
3. On the same page, click the Token link next to the key, approve access for your account, and copy the token Trello shows (ATTA…).
4. Paste both here and Connect.

Security note: The token acts as you on every board you can see. Revoke it any time from your Trello account settings (Applications).

Discord (as a tool)

Your agent posts to and reads your Discord server. Different from the Discord channel: the channel is where you TALK to your agent; this lets the agent USE Discord — post updates, read history — from any channel it's in.

1. Open the Discord Developer Portal → New Application. Then under Bot, click Reset Token and copy the token (shown once). discord.com/developers/applications
2. Under Bot → Privileged Gateway Intents, enable Message Content Intent if you want the agent to read message text.
3. Invite the bot to your server with the permissions you want (OAuth2 → URL Generator → scope bot). The agent can only do what the bot is allowed to do. Discord Docs: OAuth2 / bot invite
4. Paste the bot token here and Connect. Your agent runs the Discord MCP server with it. Discord MCP server

Security note: The bot only sees servers it's invited to and can only do what its server permissions allow — keep them minimal (e.g. read + send in a single channel). Reset the token any time from the Developer Portal.

Microsoft Teams

Your agent posts to and reads your Teams channels and chats.

Slack (as a tool)

Your agent posts to and reads your Slack workspace. Different from the Slack channel: the channel is where you TALK to your agent; this lets the agent USE Slack — post updates, read history — from any channel.

1. Create a Slack app: Create New App → From scratch → pick your workspace. (Already connected Slack as a channel? You can reuse that app and skip to step 2.) api.slack.com/apps
2. In the app's OAuth & Permissions page, add these Bot Token Scopes:
   channels:history · channels:read · chat:write · reactions:write · users:read · users:read.email
3. Click Install to Workspace (top of the same page) and approve. Copy the Bot User OAuth Token — it starts with xoxb-. Slack Docs: token types
4. Get your workspace ID: open Slack in a browser — the URL is app.slack.com/client/T…/ and that T… part is the team ID.
5. Paste both here and Connect. Then invite the bot wherever it should read or post:
   /invite @YourAppName

Security note: The bot only sees channels it has been invited to. Skip users:read.email if you don't want the agent resolving teammates' emails.

Apollo.io

Prospecting and enrichment from Apollo — reachable through the app gateway.

Attio

Records and lists from your Attio workspace — reachable through the app gateway.

GoHighLevel

Contacts, conversations and pipelines from your GHL sub-accounts — reachable through the app gateway.

HubSpot

Your agent works your HubSpot CRM — contacts, companies, deals, tickets — via HubSpot's official MCP server.

1. In HubSpot, go to Settings → Integrations → Private Apps and click Create a private app. app.hubspot.com → Settings → Private Apps
2. On the Scopes tab, grant the CRM scopes you want the agent to use — read scopes are plenty for questions; add write scopes only if it should create/update records:
   crm.objects.contacts.read · crm.objects.companies.read · crm.objects.deals.read · (optional) the matching .write scopes
3. Create the app, then copy the Access token (pat-…), paste it here and Connect. Your agent talks to HubSpot's official MCP server with it. HubSpot Docs: MCP server

Security note: A private-app token is scoped to exactly the permissions you grant — keep it read-only for a question-answering agent. Rotate or delete it any time from the same Private Apps page. (Prefer one-click? HubSpot is also reachable via the OAuth connect card above.)

Pipedrive

Deals, contacts and pipelines from Pipedrive — reachable through the app gateway.

Salesforce

Records, reports and SOQL from your Salesforce org — reachable through the app gateway.

Klaviyo

Email/SMS flows, segments and metrics — reachable through the app gateway.

Mailchimp

Audiences and campaigns from your Mailchimp account — reachable through the app gateway.

SendGrid

Transactional email sending and stats — reachable through the app gateway.

PayPal

Transactions and invoicing from your PayPal business account — reachable through the app gateway.

QuickBooks

Invoices, customers and reports from QuickBooks Online — reachable through the app gateway.

Square

Your agent reads payments, customers, catalog and orders from your Square account — read-only — via Square's official MCP server.

1. In the Square Developer Dashboard, open (or create) an application, switch to Production, and copy the Production Access Token. developer.squareup.com → Apps
2. Read more on access tokens (and OAuth, if you'd rather scope a token) in Square's docs. Square Docs: access tokens
3. Paste the production token here and Connect. Your agent runs Square's official MCP server in read-only mode (writes disabled). Square Docs: MCP server

Security note: Defense in depth: the agent always runs the Square MCP server with writes disabled, so it can read payments and catalog but never move money or edit listings. Treat the access token as a live credential and rotate it from the Developer Dashboard any time.

Stripe (as a tool)

Your agent answers questions about your Stripe account — customers, payments, subscriptions, invoices — via Stripe's official hosted MCP server.

1. In the Stripe Dashboard, open Developers → API keys and click Create restricted key — choose "Read" on the resources you want the agent to see (Customers, Charges, Subscriptions, Invoices…). dashboard.stripe.com/apikeys
2. Copy the rk_… key, paste it here and Connect. Your agent talks to Stripe's official hosted MCP server (mcp.stripe.com) with it; the tools it gets match what the key allows. Stripe Docs: MCP
3. Testing first? Use a test-mode key from the test dashboard instead. dashboard.stripe.com/test/apikeys

Security note: Strongly recommended: a RESTRICTED key (rk_…) with read-only permissions, never your full secret key. A read-only key can't move money, issue refunds or change subscriptions — and you can roll it from the same page any time.

Xero

Accounting data from your Xero organisation — reachable through the app gateway.

Shopify

Orders, products and customers from your Shopify store — reachable through the app gateway.

WooCommerce

Your agent works with your WooCommerce store — orders, products, customers — via the WooCommerce REST API.

1. In WordPress admin, open WooCommerce → Settings → Advanced → REST API and click Add key. WooCommerce Docs: REST API keys
2. Set Permissions to Read (recommended — the agent only answers questions) or Read/Write if it should update products, then Generate API key. Copy the Consumer key (ck_…) and Consumer secret (cs_…), shown once.
3. Paste your store URL, consumer key and consumer secret here, then Connect.

Security note: Prefer a Read-only key for a question-answering agent — it can't change orders, products or stock. Revoke the key any time from the same WooCommerce REST API page.

Airtable

Your agent reads and updates the Airtable bases you grant it.

1. Open Airtable's token builder and create a personal access token. airtable.com/create/tokens
2. Add these scopes (drop the write scope for read-only):
   data.records:read · data.records:write · schema.bases:read
3. Under Access, add the specific bases (or workspaces) the agent should see — it can only ever touch what you grant here.
4. Copy the token (pat…), paste it here and Connect.

Security note: Access is opt-in per base — the token sees nothing you don't explicitly grant. Revoke it from the same Airtable page any time.

MongoDB

Your agent answers questions from your MongoDB database — read-only — via MongoDB's official MCP server.

1. Get your connection string. Atlas: Database → Connect → Drivers, and copy the mongodb+srv://… URI (fill in your user/password). MongoDB Docs: connection strings
2. Strongly recommended: create a dedicated read-only database user (Atlas: Database Access → Add New Database User → built-in role "Only read any database") and use it in the string. MongoDB Atlas: database users & roles
3. Paste the full connection string here and Connect. Your agent runs MongoDB's official MCP server in --readOnly mode no matter what. MongoDB Docs: MCP server

Security note: Defense in depth: the agent always runs the MongoDB MCP server with --readOnly, so it can query and explore but never write — but the credential itself should still be least-privilege. Use a read-only user, and avoid your production primary if a secondary/analytics node exists.

Postgres

Your agent answers questions straight from your database — read-only.

1. Find your connection string. Supabase: Project Settings → Database → Connection string. Neon, RDS and friends have an equivalent page. supabase.com → Project Settings → Database
2. Strongly recommended: create a dedicated read-only role and use it in the string:
   CREATE ROLE agent_ro LOGIN PASSWORD '…'; GRANT CONNECT ON DATABASE mydb TO agent_ro; GRANT USAGE ON SCHEMA public TO agent_ro; GRANT SELECT ON ALL TABLES IN SCHEMA public TO agent_ro;
3. Paste the full postgresql://… string here and Connect.

Security note: Defense in depth: the agent's database tool runs in restricted mode (read-only transactions with execution-time limits) no matter what — but the credential itself should still be least-privilege. Never paste a superuser or table-owner connection string, and avoid your production primary if a read replica exists.

Redis

Inspect keys, queues and cache state — reachable through the app gateway.

Supabase

Your agent works with your OWN Supabase project — read-only: query tables, inspect schema, read logs.

1. Create a personal access token: Supabase → Account → Access Tokens → Generate new token. Copy it (sbp_…, shown once). supabase.com → Account → Access Tokens
2. Find your project ref (20 lowercase letters): Project Settings → General, or the subdomain of your project URL (https://<ref>.supabase.co). supabase.com → Project Settings → General
3. Paste both here and Connect. Your agent talks to Supabase's official MCP server, always in READ-ONLY mode and scoped to just this project. Supabase Docs: MCP server

Security note: Your agent runs the Supabase MCP server with --read-only and pinned to this one project, so it can explore and query but never mutate your data or touch other projects. The access token grants management-API access — revoke it any time from the same Supabase page. (Prefer a raw read-only DB connection instead? The Postgres integration takes a connection string.)

Cloudflare

DNS records, Workers and zone analytics — reachable through the app gateway.

DigitalOcean

Your agent works your DigitalOcean account — Droplets, App Platform, databases, networking and more — via DigitalOcean's official MCP server.

1. In the DigitalOcean control panel, open API → Tokens and Generate New Token. Choose scopes — pick read-only if you only want the agent answering questions. cloud.digitalocean.com → API → Tokens
2. Copy the token (dop_v1_…, shown once), paste it here and Connect. Your agent talks to DigitalOcean's official MCP server with it. DigitalOcean Docs: MCP server

Security note: The token's own scopes are the real boundary — a read-only token lets the agent inspect but never change your infrastructure. Revoke it any time from the same API page.

Netlify

Your agent manages your OWN Netlify sites — list/inspect sites and deploys, read build logs, and (with a write token) trigger deploys.

1. Create a personal access token: Netlify → User settings → Applications → Personal access tokens → New access token. Copy it (shown once). app.netlify.com → User settings → Applications
2. Paste it here and Connect. Your agent talks to Netlify's official MCP server with it; what it can do matches the token's scope. Netlify Docs: MCP server

Security note: The token acts as you on Netlify. For a read-only agent, prefer a token with the least scope you can; for deploys it needs write access. Revoke it any time from the same Netlify settings page.

Vercel

Deployments, projects and logs from Vercel — reachable through the app gateway.

Box

Search and read files in your Box account — reachable through the app gateway.

Dropbox

Search and read files in your Dropbox — reachable through the app gateway.

Google Drive

Your agent lists and reads the Drive files it creates, and can attach them to email.

OneDrive

Your agent searches and reads files in your OneDrive / SharePoint.

Freshdesk

Tickets and contacts from Freshdesk — reachable through the app gateway.

Intercom

Conversations and contacts from Intercom — reachable through the app gateway.

Zendesk

Tickets, customers and macros from Zendesk — reachable through the app gateway.

Reddit

Read subreddits and post from your account — reachable through the app gateway.

YouTube

Channel stats, video metadata and transcripts — reachable through the app gateway.

Mixpanel

Event analytics from your Mixpanel project — reachable through the app gateway.

PostHog

Product analytics, feature flags and session insights — reachable through the app gateway.

Make

Run your Make scenarios from your agent via Make's MCP server.

1. In Make, enable the MCP server (Make grid → your token) and copy the MCP URL for your region. Make Help: MCP server
2. Paste the URL here and Connect.

Security note: The Make MCP URL embeds your token — treat it as a secret. We store it write-only and you can rotate it from Make any time.

n8n

Trigger workflows on your OWN n8n instance from your agent.

1. In your n8n instance, add an MCP Server Trigger node (or a webhook), choose the workflows/tools to expose, and copy its production URL. n8n Docs: MCP Server Trigger
2. Paste the URL here and Connect. It must be reachable from the internet (your hosted n8n).

Security note: The URL points at your own n8n instance and may embed an auth token — we store it write-only. Scope the exposed workflows to only what the agent should run.

Zapier

Reach 7,000+ apps and your Zaps through your personal Zapier MCP server.

1. Open Zapier MCP, create (or open) your MCP server, and choose the actions/apps you want your agent to use. mcp.zapier.com
2. Copy your server URL (it contains your token — treat it as a secret), paste it here and Connect. Zapier Help: MCP server

Security note: The MCP URL embeds your Zapier token — anyone with it can run your actions. We store it write-only in the vault. Rotate it from the Zapier MCP dashboard any time; what your agent can do matches the actions you enabled there.