Skip to main content
Connector ID pylon The Pylon source reads support issues and their messages, customer accounts, contacts, projects, tasks, surveys, and knowledge-base content. It uses an HTTP manifest. This connector is in preview. Its requests have been checked against the official API reference, but it has not been validated against a live Pylon account. Pylon does not expose an API version selector.

Configuration

An administrator creates tokens under Settings → Integrations & dev → API tokens. Tokens use their assigned role or the current role of the user they act as. They do not expire, but deactivating the creating or acted-as user revokes them. An IP allowlist must permit your worker’s outbound address. See Pylon authentication. US connections use https://api.usepylon.com. EU connections use https://api.eu.usepylon.com. Connection testing calls GET /me. The token must be able to read every selected resource. Use organization-wide survey access for survey_responses. Pylon requires additional account and contact filters for scoped survey callers, which this connector does not supply.

Resources

Accounts, contacts, issues, users, teams, and tags are selected by default. All other resources are optional. Selecting a child also reads its parent, but only selected resources are written to the destination. Each row represents one API record. IDs are strings. Child tables carry the parent ID in their key, and followers also include their user or contact type. Status tables use their stable slug. Nested custom fields, relationships, SLA data, survey answers, form definitions, macro actions, training documents, and recording segments remain JSON columns. Every resource includes a raw column for unmapped fields. Issue search has no creation-time window, so issues reads historical issues without the 365-day window required by GET /issues. Messages include internal notes and replies. Threads and followers are separate tables. issue_voice_calls scans issues and requests calls only for issues whose source is phone. It preserves transcript segments and audio links inside each call’s recordings. call_recordings reads the separate meeting-recording metadata API. Articles use each knowledge base’s default language and include both draft and published HTML and Markdown, with embedded media retained. Collections preserve their parent collection ID. active_themes reads the active theme of each knowledge base, including draft and published asset contents. Article feedback is read across knowledge bases.

Coverage limits

  • Feature requests. feature_requests returns at most 1,000 records without their evidence. Pylon’s schema exposes a limit but no cursor for this endpoint, despite its description mentioning pagination. It is not a complete backfill for larger accounts. See the feature request reference.
  • Lookups by known ID. Pylon cannot list custom object types, notebook blocks, or inactive themes, so those are not replicated. opportunities uses the documented opportunity search alias. Its additional opportunity attributes are preserved in raw.
  • Translations. Articles are read in each knowledge base’s default language only. Other language codes listed in knowledge_bases.supported_languages are not read.
  • Milestones. Pylon excludes milestones belonging to unlaunched projects from search. Those are not discovered by this connector.
  • Recordings and files. Phone recordings appear only after their transcripts finish processing. Meeting recordings contain metadata and summaries, not a downloadable transcript body. Attachment and audio URLs are preserved, but binary files are not downloaded. Phone audio URLs expire.
  • Visibility. Reads reflect the token’s role and available product features. Users include deactivated users. Removed records and records hidden by permissions are not available as deletion events.
The connector uses bulk list or search endpoints when their response includes the same record as a detail endpoint. It does not duplicate those reads. Account files, account highlights, attachment uploads, and message edits expose write operations rather than independent read collections. They are not separate resources. Audit logs are excluded. See the full API reference for the upstream surface.

Modes

All resources perform full reads. Incremental reads are not advertised. Changes to nested messages, evidence, transcripts, and related records do not have a verified common update cursor. Use full upsert to refresh rows, or full replace to remove rows no longer returned within the selected scope. Do not use the capped feature_requests as a complete replacement for a feature-request dataset.

Behavior

  • Auth. Every request sends the token as a Bearer credential to the selected regional API host. Authentication and permission errors fail the read.
  • Pagination. Paginated endpoints follow pagination.cursor while pagination.has_next_page is true. GET requests put the cursor in the query and POST searches put it in the JSON body. Pages contain 100 records, except account email activities, which use 20. Endpoints documented to return all records without a cursor are read once. Feature-request search is the capped exception described above.
  • Rate limiting. Requests are paced at two per second with a burst of one, matching the 120-per-minute search allowance. Reference endpoints with a 30-per-minute allowance are fetched once per run. Pylon’s allowances are per organization and endpoint, shared with other clients. The HTTP engine retries throttled requests and honors Retry-After.
  • Recovery. Top-level paginated reads can resume from saved page cursors. Child reads restart from their parent listing. Full upsert avoids duplicate rows when replaying.
  • Parent selection. Optional child reads may require many requests. Phone calls scan issue search separately to select phone issues. Knowledge bases without an active theme do not produce an active-theme request.
See Pylon API conventions and rate limits for pagination and service limits.