Integrations
Connect a bot to Google Sheets, Drive, GetCourse, or AI. Where to save access and which reaction action to add.
บนหน้านี้
An integration lets your bot work with another service, for example adding an inquiry to Google Sheets or sending a customer to GetCourse. First save the connection, then select it in a reaction action.
Open Integrations → Add integration, choose a service, and fill in its fields. Access details are stored encrypted. On the connection card, you can turn it on or off or open Configure.
Available services
- AI: a language model (OpenAI-compatible API). See AI answers and agents.
- Google Sheets: appending, reading, finding and updating rows by column headers (see Actions).
- Google Drive: uploading files to a folder.
- GetCourse: creating and updating users and deals, exporting users, deals and groups.
- Webhook: send data to another service's address over HTTP.
- External handler: hand off a reaction to your own program on a server through a persistent WebSocket connection. See External handler.
Google Calendar is also a connected service, but it is not added from this screen: connect it with OAuth in AI → Calendar, where meeting types are configured as well. See Meetings and Google Calendar.
For CRM systems such as amoCRM/Kommo, Bitrix24, RetailCRM, and OzmaCRM, use Integrations → CRM. There you can connect the system, match its fields with GetMyBot fields, preview the result, inspect the sync log, and resolve conflicts. Data syncs in both directions. amoCRM/Kommo use OAuth, RetailCRM an API key, OzmaCRM OIDC, and Bitrix24 an incoming portal webhook. See CRM integrations for setup steps.
Configuration
When adding an integration, choose the Service, give the connection a Name, and fill in the fields for that service:
Webhook
- Webhook URL: the receiver's address.
- HMAC signature (secret): a shared key the receiving service uses to confirm the request came from GetMyBot. The signature is sent in the
X-Signature: sha256=...header.
GetCourse
- Account (the school name, for example
myschool) and the API Secret key. - Object: "Users" or "Deals".
The GetCourse action does two kinds of work.
Import into GetCourse. Fields are grouped as in GetCourse's official API: User (email, phone, first_name, last_name, city, country, group_name), Import settings (for example refresh_if_exists), Traffic source (utm_source, utm_campaign, referer), Deal (offer_code, deal_cost, deal_status, deal_is_paid), and Custom user fields. A value can come from customer data, such as email = {{param.email}}, or be entered directly, such as utm_source = telegram. Group and custom-field suggestions come from your account; if they cannot load, enter names manually. Existing actions with one flat field list still work. Convert to sections puts known keys in their sections and the rest in custom fields.
If GetCourse responds with HTTP 200 but says success: false, the action fails and shows error_message. On success, {user_id, deal_id, lead_id} are saved in the reaction's temporary data; missing values remain zero.
Get data from GetCourse. Export users, Export deals, and Export groups start an export in GetCourse and wait up to 25 seconds for it. If it is not ready, the action fails with export_not_ready: retry later or narrow the filter. A filter consists of "key, type, value" rows (for example created_at[from] = 2026-10-01); unknown keys are passed through unchanged. The result is a list of records with the account's field names, at most 1000 rows per run (default 50). has_more under Page details shows whether another page exists. GetCourse limits export frequency, so do not run an export for every message. Use infrequent scenarios or a schedule.
Google Sheets
- Spreadsheet link (or its ID).
- Service account JSON (secret): a file with the key for a separate Google account used by software. Share the spreadsheet with that account's email address. Otherwise reading and writing fail with an access error, shown in the action editor and during the check.
Action fields are matched to the column headers of the header row (the first row by default, another one can be set). Operations:
- Append a row: every field is written to the column with the same header, so the column order of the sheet does not matter. An unknown header is an error. Result:
{row_number, range}. - Read rows: the rows after the headers as objects plus the
_rownumber. The default page is 50, the maximum 500;has_moreandnext_offsetcan be saved under a separate key. - Read the last row and Find a row: scan the sheet page by page and fail if the sheet is too large, instead of returning a wrong row.
- Update a row: changes the cells of row
row_number(for example{{ctx.found.row_number}}after a search).
Values can be written "as typed" (Google Sheets recognizes formulas and dates) or "as text" (without recognition). The editor offers column headers as field names; if they could not be read, enter names manually. Preview rows shows up to 20 rows and changes nothing in the sheet. You can use the preview as a sample for response parsing, which selects values for later actions.
Google Drive
- Folder ID.
- Service account JSON (secret). Give that account editor access to the folder.
The Google Drive action uploads a file with a name, file type (MIME), and content. It saves {file_id, web_view_link} in the reaction's temporary data, for example under upload. You can then insert the link with {{ctx.upload.web_view_link}}. As with other actions, you can configure response parsing.
AI
- Base URL, Model, System prompt, API key (secret). See AI answers and agents for details.
One AI connection is used by the AI action in reactions, the AI agent, and the preparation of knowledge sources for search. For search, the model creates numerical representations of text called embeddings. They use the same base URL and key, with the text-embedding-3-small model by default. Without a working AI connection, sources cannot be prepared for search and the agent cannot be activated.
External handler
A WebSocket integration: your server connects to GetMyBot and processes reactions by responding with Telegram Bot API commands.
- WebSocket address: the address your service connects to.
- Connection token (secret): shown once at creation; a reissue button is available.
Details, protocol, and service examples: External Handler.
Using integrations in reactions
- AI and Webhook are available as actions/channels immediately: add an AI action or send data using the Web request action to the webhook address.
- The GetCourse, Google Sheets and Google Drive actions are available in the reaction builder. Each has optional response parsing: pick values from the action result and put them into the context; a failure runs the configured error route. If the feature is not in the plan, the step is marked failed with the reason
feature_unavailable.
Checking an integration
Each integration card has a Check button. It writes nothing: for Google Sheets it reads spreadsheet metadata, for Google Drive folder metadata, for GetCourse it makes a safe read call, and for AI it only validates that the settings are filled in (no paid call). There is nothing to check for other types. The result is shown per check: what passed and what did not, with a short explanation.
Security
GetMyBot does not send integration requests to internal networks and limits response size. This protects addresses not intended for outside access (SSRF). Integration secrets are stored encrypted.
Knowledge crawling and file extraction add their own limits on top of that: only allowlisted HTTPS origins, robots.txt compliance, bounded redirects, size and time caps, and no execution of macros, scripts, or embedded objects. See Knowledge sources.
What's next
- AI answers and agents: configuring and using the model.
- Knowledge sources: what the AI agent may answer from.
- Meetings and Google Calendar: connecting a calendar.
- External Handler: connecting your own backend via WebSocket.
- CRM integrations: amoCRM/Kommo, Bitrix24, RetailCRM, and OzmaCRM.
- Sources and Keys: REST/OzmaDB connectors for reading and writing data.
- Web Requests and Webhooks: inbound and outbound HTTP.
Speech recognition
Choose an OpenAI-compatible provider or a custom service, then set its base API URL, model and API key. An optional language code, such as en or ru, hints the speech language; leaving it empty lets the provider detect the language. The key and model must belong to the configured provider.