> For the complete documentation index, see [llms.txt](https://developer.celigo.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://developer.celigo.com/changelog/2026.md). # 2026 {% updates format="full" %} {% update date="2026-07-20" tags="mcp,feature" %} ## OAuth sign-in for Platform MCP Add the [Platform MCP](https://developer.celigo.com/mcp) endpoint to your client with no credentials and it opens a browser sign-in — no more creating and copying API tokens. Per-client setup in [Connect a client](https://developer.celigo.com/mcp/connect). ```json { "mcpServers": { "celigo": { "url": "https://api.integrator.io/celigo-mcp" } } } ``` * The server implements OAuth 2.1 with PKCE and dynamic client registration. Clients that follow the MCP authorization spec discover the sign-in automatically — current releases of Claude, Cursor, VS Code, Windsurf, and ChatGPT all do. * Sign-in uses your Celigo credentials, including SSO and MFA where your account requires them. The session acts as you, in the account and environment you pick at sign-in. * Clients refresh the session automatically. You sign in again only if access is revoked, you switch account or environment, or the session goes unused for an extended period. * API tokens and personal access tokens still work for CI, scripts, and clients without OAuth support — pass one in the `Authorization` header, same as before. * The endpoint is region-specific: `https://api.eu.integrator.io/celigo-mcp` signs in against `eu.integrator.io`. To switch account or environment, disconnect the server and sign in again. {% endupdate %} {% update date="2026-07-14" tags="api,feature" %} ## Personal access tokens A personal access token (PAT) is a bearer token you create yourself that acts with your own user permissions — no admin provisioning, no scopes. Full details in [Authentication](https://developer.celigo.com/api/getting-started/authentication). * Create one under **Resources → API tokens** → **+ Generate token** → **Personal access token**. Manage and Monitor roles get self-service API access for the first time; for them it is the only token type offered. * The token inherits your permissions and follows role changes. Disabling your user blocks the token; deleting your user revokes it. * Admins can see and revoke every PAT on the account, but cannot view a token value they didn't create. * PATs auto-purge after 90 days by default, configurable from 1 hour to never. * A PAT works anywhere an API token works: REST API calls, `CELIGO_API_TOKEN` in the CLI, and the `Authorization` header on Platform MCP. {% endupdate %} {% update date="2026-07-14" tags="mcp,feature" %} ## Platform MCP: nine new tools Connected clients pick these up automatically — the catalog is fetched at runtime. Full details in [Tools](https://developer.celigo.com/mcp/tools). **B2B EDI** * `list_edi_profiles` / `upsert_edi_profile` manage trading-partner interchange envelopes (X12 ISA/GS, EDIFACT UNB). * `list_file_definitions` / `upsert_file_definition` manage parsing and generation rules for CSV, fixed-width, X12, and EDIFACT files. * `list_edi_transactions` queries the EDI transaction log by document type, direction, document number, or date window; `includeFaDetails` / `includeMdn` add acknowledgment detail. * `update_edi_fa_status` marks acknowledgments `accepted` or `rejected` in bulk. **Marketplace and debugging** * `list_marketplace` browses published templates or previews one template's full blueprint; `deploy_template` installs it as a new integration mapped to your connections. * `list_execution_logs` reads a run's debug logs: the index, one record's step timeline, or request/response payloads per stage. Requires debug logging armed on the flow. **Existing tools** * `list_flow_errors` with no arguments returns an account-wide summary of every erroring flow and step. * `list_flows` adds `hasOpenErrors`, `includeErrorCounts`, and `numError_gte`. * `run_flow` takes optional overrides: `export.startDate` / `export.endDate` for backfills, `_exportIds` to run a subset of source exports. * `list_jobs` with `includeFiles: true` returns short-lived download URLs for files a run produced. **Fixes** * Every tool declares MCP tool annotations (read-only, destructive, idempotent). * `run_flow` and `cancel_job` take `_id`, matching every other tool (previously `id`). * `update_flow_error_retry_data` advertises the correct body schema — the retry-data envelope, not the flow schema. `list_lookup_cache_data` drops a stray `body` parameter. * `delete_resource` states that its dependency pre-flight runs for connections only and that warnings never block the delete; `resourceType` adds `edi-profiles` and `file-definitions`. * `list_environments` returns an empty list with a warning on accounts without the feature. `list_jobs` returns `[]` instead of `null` when nothing matches. {% endupdate %} {% update date="2026-07-14" tags="api,feature" %} ## July API updates Everything below is live in production and documented in the [API reference](https://developer.celigo.com/api). **AI agents** * AI agent and guardrail imports support Anthropic Claude: set `provider: "anthropic"` with a Claude model id in `litellm.model` and Claude settings in `litellm._overrides.anthropic` (required `systemInstruction`, extended thinking via `thinkingConfig`, Claude tools including web search). `temperature` is capped at 1 and `responseFormat.type` must be `text` or `json_schema`. * All three providers accept a `resources` array — read-only reference content (policies, schemas, documentation) pinned from connected MCP servers. * OpenAI reasoning effort adds `none` and `xhigh`; OpenAI configs support `topLogprobs`; the LiteLLM `maxCompletionTokens` default rose from 1000 to 5000. **Endpoints** * The Syncs reference now covers the full build surface, not just running and monitoring: sync CRUD, dataset CRUD, schema-drift events, source and destination metadata catalogs, and usage endpoints. * `POST /v1/connections` accepts `__integrationIds` (up to 100) to create a connection and register it on integrations in one call; post-save registration failures are reported in `__failedIntegrationRegistrations` without rolling back the connection. * The step error endpoints (`GET /v1/flows/{_id}/{_stepId}/errors` and `.../resolved`) add filters: `_flowJobId` (paired with `occurredAt_gte` or `resolvedAt_gte`), occurred/resolved date bounds, `tags` (including `untagged`), and `resolvedBy` (including `auto`). * `GET /v1/apis/{_id}/template` and `GET /v1/tools/{_id}/template` export a builder-mode API or Tool and everything it references as an installable `.zip` via a short-lived signed URL. * `POST /v1/apis/{_id}/clone/validate` dry-runs a clone and returns `canClone`, catching version/route conflicts before anything is created. * MCP servers (the `mcpservers` resource) accept a `resources` array: Celigo Storage files served to MCP clients through `resources/list` and `resources/read`. **Also in the reference** * Integration clone previews include Tool resources; the `tools_not_supported` rejection is gone. * The recycle bin supports the `storageitems` resource type (list-only). * License responses document `storage` entitlements (`maxAllowedUsage`, `disableOverage`) and EDI `labelPrinting` entitlements. * Trading-partner connectors carry new `type` (`FTP`, `AS2`, `S3`) and `region` fields. {% endupdate %} {% update date="2026-06-18" tags="mcp,feature" %} ## Celigo Platform MCP is generally available The Celigo Platform MCP is now live at `https://api.integrator.io/celigo-mcp`. It's a hosted remote MCP endpoint that lets AI agents read, run, and troubleshoot your integrator.io account: no local process, no sidecar, a bearer token and any MCP-compatible client. ### Who it's for Anyone working with AI coding agents: Claude Code, Claude Desktop, Cursor, VS Code with Copilot, Windsurf, or any client that speaks Streamable HTTP. Point the client at the endpoint, hand it an API token, and the agent can operate on your integrator.io account directly. ### What agents can do today * **Inspect everything.** List and get connections, integrations, flows, exports, imports, scripts, agents, guardrails, APIs, tools, MCP servers, lookup caches, iClients, jobs, and audit entries. * **Build and edit resources.** Create or update any resource with a single `upsert_*` tool, and delete any resource with `delete_resource`, which reports dependents as advisory warnings but doesn't block. A masked-credential guard keeps agents from overwriting stored secrets. * **Run flows and triage errors.** Trigger a flow, wait for the job, inspect per-step results, then retry, resolve, tag, or assign failed records, all in one conversation. * **Use built-in prompts.** Guided workflows ship with the server: account health audits, flow troubleshooting, connection diagnosis, integration planning, flow config review, and Handlebars/SQL writing assistance. The `search_knowledge_base` tool answers product questions straight from the Celigo Knowledge Base. * **Load reference resources.** The server exposes a product glossary, tool usage guide, error pattern reference, connector catalog, and API reference as MCP resources: context the agent pulls on demand. (Full connector and resource schemas come from the `get_schema` tool, not a resource.) ### Get started ```json { "mcpServers": { "celigo": { "url": "https://api.integrator.io/celigo-mcp", "headers": { "Authorization": "Bearer " } } } } ``` Full setup, authentication details, transport options, and the complete tool catalog live in the [Platform MCP section](https://developer.celigo.com/mcp).
Notes * Requires a full-access API token * Region-aware: use `https://api.eu.integrator.io/celigo-mcp` for EU accounts
{% endupdate %} {% update date="2026-06-17" tags="cli,feature" %} ## CLI 2026.6.1 `@celigo/celigo-cli` **2026.6.1** is on npm. A large release: seven new command groups, file-based input across every write command, credential-safe `set`, actionable error hints, and a round of command renames that align the CLI surface. ```bash npm install -g @celigo/celigo-cli ``` ### New command groups * `celigo connectors`: base CRUD plus `install-base`, `install`, `push-update`, and `connectors licenses` CRUD. * `celigo state`: read and write resource and account state with `list`, `get`, `set`, `delete`, `purge`. Global by default; scope to a resource with `--resource-type` and `--resource-id`. * `celigo syncs`: `run`, `audit`, `sources`, `destinations`, `cancel-jobs`. * `celigo sync-jobs`: `list` (auto-paginated per sync, newest first, with `--created-gte` / `--created-lte`), `get`, `errors`, `cancel`. * `celigo recycle-bin`: `list`, `get`, `restore` (with `--cascade`), `purge`. * `celigo mcp-oauth-providers`: full CRUD for MCP OAuth providers. * `celigo event-reports`: `list`, `get`, `create`, `signed-url`, `cancel`. ### New subcommands on existing groups * HTTP connectors: `create`, `update`, `delete`, `resources`, `resource`, `endpoints`, `endpoint`. * Templates: `templates install` to install from a local `.zip` or by id with a connection map, plus `list`, `get`, `create`, `update`, `delete`. * EDI transactions: `fa-detail`, `mdn-detail`, `update-fa-status`, `download-file --document-type`. * APIs: `set-group` / `unset-group`, plus `create-api-group` / `delete-api-group` / `api-groups`. * Flows: `error-summary`, `delete-execution-logs --started-at --end-at`, `cancel-jobs`, `set-group` / `unset-group`. * Connections: `applications` (external apps in use, filterable with `--application`), `delete-debug-logs`, `purge-messages`. * Scripts: `audit`, `delete-debug-logs`. * Tools and partners: `tools connections`, `tp-connectors create/update/delete`. * Across resources: `dependencies` (alias `used-by`), `audit`, `profile whoami`, `on-premise-agents installer-url --os`, `notifications list`, `stacks audit`, `users enable/disable`, `environments enable/disable`. ### File-based input `-f, --file ` now works on every `create`, `update`, and `install` command (`--file -` still means stdin). And `set` accepts `file://` assignment values to load multi-line content such as script source or SQL without escaping: ```bash celigo exports create -f ./export.json celigo scripts set 64a... content=file://./transform.js ``` This avoids shell input redirection, which is unreliable when driven by an agent and unsupported in PowerShell. ### Credential-safe writes * `set` uses atomic JSON-Patch whitelists. Whitelisted fields PATCH cleanly; masked-credential endpoints (connections, iClients) refuse non-whitelisted edits rather than round-tripping a GET then PUT that would wipe stored secrets. * `update` refuses payloads that still contain the masked `******` placeholder unless you pass `--force`. * `profile delete` now confirms before deleting (`-y, --yes` to skip). ### Better errors * Actionable hints mapped from API error codes (`inactive_flow`, `feature_not_enabled`, `invalid_ref`, `access_restricted`) now appear in the formatted error. * Empty required positional arguments are rejected up front, catching unset shell variables like `celigo jobs get "$ID"` before they hit the API blind. * `--limit` is validated as a positive integer instead of silently coercing bad input. ### Renamed and restructured commands * `account deps` is now `account dependencies`. * ` debug-enable` / `debug-disable` are now `enable-debug` / `disable-debug`. * `exports` / `imports` `test-run-step` is now `test-run-step-results`. * `flows flow-group-create` is now `create-flow-group`; `flow-group-assign` is now `set-group` (with a new `unset-group`). * `flows execution-logs-enable` / `-disable` are now `enable-execution-logs` / `disable-execution-logs`. * `flows execution-log-query` is now `query-execution-logs`; `execution-log-data` is now `execution-log-detail`. * `integrations snapshot` is now `create-snapshot`; `integrations files` plus `download` are now `download-files`. * `agents` / `scripts` `logs` is now `debug-logs`. * `usage historical` is now `historical-usage`. * Token verbs consolidated: `system-token` / `display-token` to `token`; `recycle-token` / `change-token` to `rotate-token`. * Error commands now take explicit ids: `resolve-errors` is now `error` (flow, step, and error ids passed as arguments), `error-analyze` is now `error-analysis`, `update-error-data` takes the error id explicitly, and `assign-errors` gained an explicit step argument. * Templates: `templates list` (marketplace) is now `templates marketplace`; `invoke` and `preview` take an optional id. * The interactive REPL was removed; running `celigo` with no arguments no longer opens a shell. * The `account-info` group was removed; its `applications` listing is now `connections applications`. ### Fixes and internals * `connections purge-messages` treats the API's "queue already empty" `404 Not Found` as a successful no-op. * `users` and `environments` `enable/disable` read current state before toggling, so the verbs are honest and idempotent. * `account dependencies` orphaned-script detection walks `_scriptId` references recursively, catching flow-level and nested refs. * New `createdAt` cursor pagination with a truncation warning when the safety cap is hit (used by `sync-jobs list`). * The account index staleness window tightened from 4 hours to 15 minutes. * All 124 bundled skill schemas re-synced from the platform's published API specs, the drift check is now line-ending tolerant, and the legacy `rest` / `simple` schemas were removed (use `HTTPExport` / `HTTPImport` instead). {% endupdate %} {% update date="2026-04-05" tags="cli,feature" %} ## CLI 2026.4.5 — generally available `@celigo/celigo-cli` **2026.4.5** is now generally available on npm. This is the launch release: a first-party command-line tool for driving integrator.io from your terminal, your scripts, your CI pipelines, and the AI coding agents you already work with. ### Who it's for Developers and AI coding agents who live in a terminal. If the integrator.io UI is the right tool for your day-to-day, keep using it — the CLI exists for the workflows where a UI slows you down: scripted bulk edits, version-controlled configuration, CI checks, and long-horizon agents that need to read and write live integrator.io state without burning context on a GUI. ### What you can do * **Drive every resource.** Anything in the REST API maps to a `celigo ` command — integrations, flows, connections, exports, imports, scripts, agents, guardrails, EDI profiles, on-premise agents, and more. * **Manage multiple accounts.** Profiles switch between prod, sandbox, and EU regions without juggling tokens. `celigo profile use eu` is the whole context switch. * **Lint your account.** `celigo account lint` runs static analysis across integrations, flows, and scripts — broken references, missing required fields, common script errors — before you push to production. * **Snapshot, search, diff.** `celigo account snapshot` pulls every resource into a local index for full-text search (`celigo account search`), dependency graphs (`celigo account deps`), and per-resource stats. * **Pipe-friendly output.** Every command supports `--format json|yaml|table|csv` and `--jq`, so results flow cleanly into shell scripts and other tools. ### AI assistant skills, included Installing the CLI globally also installs a set of AI assistant skills for Claude Code, Cursor, and Windsurf — domain-specific guides plus OpenAPI-derived schemas the assistant loads on demand. Ask "build a Shopify → NetSuite flow" and the assistant pulls the `building-flows` skill and writes JSON that actually validates against integrator.io, instead of guessing field names. Skills update atomically when you upgrade the CLI. ### Get started ```bash npm install -g @celigo/celigo-cli celigo profile add prod --api-token "$CELIGO_API_TOKEN" celigo integrations list --format table ``` Full install options, profile setup, region targeting, and skills coverage live in the [CLI section](https://developer.celigo.com/cli).
Notes * Requires Node.js 22+ * MIT licensed * Region-aware — pass `--api-base-url https://api.eu.integrator.io` for EU accounts, or scope it per profile
{% endupdate %} {% endupdates %} --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://developer.celigo.com/changelog/2026.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.