How to Connect Every DocsKoala Integration: The Full Setup Guide

• Drafting sources (GitHub, Zendesk, Intercom, Help Scout, the Custom API webhook, Site Import): these feed the AI drafting pipeline. A merged PR, a solved ticket, or a scraped page becomes a candidate draft, and every one of them lands in the Review Inbox for a human to approve. None of them publish anything on their own.
• Inbox channels (Slack, Telegram, Email, the Custom API Channel): these are two-way. A customer message on that channel becomes a conversation in your Inbox, and a reply you send from DocsKoala goes back out over the same channel, so the customer never knows they're talking to a support tool with a dashboard behind it.
• Context and action connectors (Linear, Shopify): these don't create conversations or drafts by themselves. Linear lets an agent create or link an issue from an open conversation. Shopify shows a customer's recent orders next to a conversation, looked up live by their email, read-only.
• Developer surfaces (MCP, the public REST API, outbound Webhooks): these let your own code or an AI coding agent read and act on the same data, gated behind the Scale plan's mcpServer/publicApi features. See the dedicated MCP setup guide for that walkthrough specifically.

GitHub is the only connector included on every plan, since it's the core PR-to-docs pipeline the product is built around. Reads merged pull request diffs from tracked repos and drafts new articles or article updates from customer-facing changes.

• 1.In Integrations, click Connect GitHub. You're redirected to GitHub's own App installation screen, not a token-paste form.
• 2.Choose which repositories to grant the DocsKoala GitHub App access to, on GitHub's side. You can select specific repos rather than your whole organization.
• 3.You're redirected back to DocsKoala. Every repo the installation can see is tracked automatically, no separate "add repo" step needed for repos included at install time.
• 4.Add repos later with the Add repo picker on the Integrations page, or manage the installation itself (add/remove repos, uninstall) from GitHub's own Manage on GitHub link.

All three do the same job: recently closed support conversations get evaluated as potential article material, on a 15-minute polling cycle (none of these use a webhook), and only public/customer-visible messages are ever sent to the drafting pipeline, internal agent notes are filtered out before anything leaves the connector.

Zendesk

• 1.In Zendesk, go to Admin Center → Apps and integrations → APIs → Zendesk API and enable token access, then generate a new API token.
• 2.In DocsKoala's Zendesk connect form, enter your subdomain (the part before .zendesk.com), your agent email, and the API token you just generated.
• 3.DocsKoala verifies the credentials against your Zendesk account before saving anything, if the subdomain or token is wrong, you'll see an error immediately rather than a silent failure later.

Intercom

• 1.In Intercom, go to Settings → Personal Access Tokens and create one.
• 2.Paste it into DocsKoala's single Access token field.
• 3.Connecting again with a new token replaces the existing connection, there's one Intercom connector per org.

Help Scout

• 1.In Help Scout, go to My Apps → Create My App to register an OAuth app and get a client ID and client secret.
• 2.Enter both into DocsKoala's Help Scout connect form.
• 3.This uses OAuth2's client-credentials grant, there's no browser redirect or login screen, DocsKoala exchanges the credentials for an access token directly.

DocsKoala has two separate "custom API" connectors that solve different problems, worth telling apart before you set either one up.

Custom API webhook (drafting source)

• For any support tool that isn't Zendesk, Intercom, or Help Scout: click Generate webhook and DocsKoala gives you a URL and a signing secret, no fields to fill in on your end.
• Your tool POSTs closed tickets to that URL as { externalId, title, body, status, url?, tags? }, signed with the secret using HMAC-SHA256 in an X-DocsKoala-Signature header. Only status: "closed" payloads trigger drafting.
• You can create more than one of these, useful if you have separate tools feeding separate projects.

Custom API Channel (two-way Inbox channel)

• For piping a support tool's live conversations into your Inbox with real replies going back out: enter your reply-webhook URL, the endpoint DocsKoala should POST agent replies to.
• DocsKoala then gives you its own inbound webhook URL and signing secret for your tool to send incoming messages to.
• Both directions are HMAC-signed the same way as the drafting webhook above.

Email (IMAP/SMTP)

• 1.This connects a specific team member's mailbox, not an org-wide inbox, each person who wants to handle email support connects their own.
• 2.Enter the email address, then incoming (IMAP) host/port/username/password and outgoing (SMTP) host/port/username/password. Leave "Use TLS/SSL" checked unless your provider specifically says otherwise.
• 3.For Gmail, generate an App Password in your Google Account security settings rather than using your real password (required once 2-factor auth is on, and better practice regardless).
• 4.DocsKoala opens a real IMAP and SMTP connection to verify the credentials before saving them, encrypted, at rest.

Telegram

• 1.Message @BotFather in Telegram, create a new bot, and copy the bot token it gives you.
• 2.Paste that token into DocsKoala's Telegram connect form.
• 3.DocsKoala verifies the token, then registers itself as that bot's webhook automatically, there's nothing further to configure in Telegram's side.

• 1.Click Connect Slack. You're sent to Slack's own OAuth consent screen, requesting permission to read and post in channels.
• 2.Approve the install in Slack, then invite the DocsKoala bot to whichever channel you want connected, if it doesn't show up in the next step, that's usually why.
• 3.Back in DocsKoala, pick that one channel from the picker. Messages posted there become Inbox conversations, and replies you send from the Inbox post back into the same thread.

• 1.Click Connect Linear and approve the OAuth consent screen. No further setup step is needed after that.
• 2.From an open conversation in the Inbox, an agent can create a new Linear issue or link an existing one directly, DocsKoala looks up any already-linked issue live rather than storing its own copy of the link.

Status changes made in Linear don't sync back automatically today, if an issue gets closed in Linear, that won't update anything on the DocsKoala side.

• 1.Enter your shop domain (just the part before .myshopify.com) and submit.
• 2.You're redirected to Shopify's OAuth consent screen for your store specifically.
• 3.Once approved, any conversation whose reporter email matches a real Shopify customer shows their recent orders alongside it, looked up live, nothing cached.

This connector is read-only by design: it shows order history with a link to open the order in Shopify admin, it can't modify orders, issue refunds, or take any action in your store.

Point DocsKoala at an existing docs or help site and it crawls same-origin pages to draft initial articles. Covered in full detail, including exactly what the crawler will and won't touch, in the dedicated Site Import guide.

All three of these are built on the same underlying object: a scoped API token, created from the MCP page, with an explicit checklist of scopes and an expiry. The token works identically whether it's used by an MCP client (Claude Code, Cursor) or a script calling the REST API directly.

• MCP lets an AI coding agent list the review inbox, draft/edit articles, and reorganize categories, everything still lands in the Review Inbox for approval. Full setup walkthrough: how to connect Claude Code or Cursor via MCP.
• The public REST API (/api/v1/*) uses the same bearer tokens: list and reply to conversations, list published articles (never unpublished drafts, even with the right scope, that's deliberate).
• Outbound webhooks, set up separately under Settings → Webhooks, notify your own endpoint when a draft is created, an article publishes, or an escalation's status changes. Each endpoint gets its own signing secret, shown once, and deliveries retry with backoff rather than firing once and giving up.