Heads up
This is a pre-release list of features for Celigo platform version 2026.4.1 scheduled for incremental release April 13 – 16, 2026.
These release notes are still being updated, and the links below may lead to unpublished articles until they are finalized on the day of the release.
Agent builder lets you build low-code AI agents that are attached to a model, given a clear set of instructions, and empowered to use their own judgement to invoke tools and APIs — without manual orchestration.
Builders can now resume previous conversations with agents to continue work without starting over.
-
Start a new chat or pick up where you left off with saved conversation history per agent
-
See live tool execution progress directly in the chat, including failures with clear next steps
-
Debug mode to inspect tool/prompt/resource traces in expandable sections
-
Attach files (images/PDFs) and send structured JSON to the agent
MCP subscription entitlements and usage are now visible and trackable directly within the platform, eliminating the need to monitor consumption separately.
MCP server builder now supports prompts, in addition to tools. You can publish reusable prompt templates through MCP, so compatible clients can discover and use the same prompt definitions. Prompts help teams standardize high-quality patterns across agents without duplicating content in every integration.
-
First-class prompt publishing: Define, and publish prompt templates from MCP Server Builder alongside your existing tool definitions
-
Client discovery and reuse: Prompts are exposed via MCP so compatible clients can discover available prompts and invoke consistent prompt definitions across agents
-
Auditability: Track prompt usage for compliance and troubleshooting
Celigo Ora is coming to beta in this release — an AI copilot that operates inside the product. Not a chatbot or doc search, but an agent that sees what you see, understands what you’re working on, and does the work. You describe what you want in plain language, and Ora does it.
-
Build real flows from plain language, with exports, imports, mappings, and schedules wired to your connections
-
Map blast radius across your entire account before you make a change
-
Diagnose errors by reading configs, inspecting HTTP traces, and correlating with history – then telling you what to fix and why
-
Perform bulk actions across flows and steps in one shot
-
Navigate the product and surface answers that aren’t directly accessible through menu options
-
Cover all 20 resource types: integrations, flows, connections, exports, imports, scripts, APIs, EDI profiles, MCP servers, users, and more
-
Write and edit code: JavaScript hooks, SQL queries, handlebars templates, field mappings, and filter expressions
Celigo tools let you build reusable integration logic once and use it everywhere – across AI agents, MCP servers, and other tools.
Tools are reusable units of integration logic with clearly defined inputs and outputs. They help eliminate duplication, improve consistency, and provide a shared foundation for scalable automation and AI-driven workflows.
Tool builder is a low-code, Flow-Builder-like experience for creating tools. Define tool logic visually using familiar constructs such as imports, lookups, branching, and transformations.
Each tool defines a clear contract through input and output schemas, created using JSON Schema and mapped using the mapper. Tools can be validated using mock input data before deployment, making them easy to test, reuse, and compose.
MCP server exposes APIs and integrations to AI agents with governed access, scoped security, and audit logging.
Tools can be used in different Celigo experiences depending on the integration logic.
-
In AI agents and MCP servers, tools are attached directly to enable agent- and server-driven behavior.
-
In other tools, tools are invoked using the tool step, allowing packaged logic to be composed as part of a larger workflow.
Support for using tools directly within flows and APIs will be introduced in a future release.
The Tool step provides deterministic execution of tool logic with structured inputs and outputs. Inputs and outputs are validated against the tool’s schemas to ensure predictable behavior across integrations.
Connections can be overridden at runtime, allowing the same tool to be reused across environments without modification. A built-in Run/Preview experience enables safe iteration during configuration.
Execution logging is now available for AI agents and guardrails, providing an end-to-end trace of how each execution processes records across your agentic workflow. For AI agent steps, the log captures both how the agent reasoned and when it took action including inputs, key decision points, actions performed, and resulting record changes, so you can quickly understand what happened and why.
For guardrail steps, execution logging records evaluation and outcomes, making it clear when records are allowed, blocked, or modified. Together, this unified visibility simplifies troubleshooting and auditing by making record-level processing and outcomes easy to trace from start to finish.
We're excited to announce the general availability of the Celigo MCP Server, the standardized way to expose your Celigo APIs, integrations, and logic to AI Agents securely, using the Model Context Protocol (MCP). The MCP Server transforms your internal automations into a stable, scalable foundation for AI-powered flows, enabling your AI assistants and agents to reliably discover, understand, and invoke your business capabilities with clarity and control.
-
Access thousands of apps: Power AI assistants and agents with governed access to thousands of apps by leveraging the comprehensive library of Celigo prebuilt connectors.
-
Expose custom business capabilities: Because every business operates differently, you can expose APIs and contextual business logic (built in API Builder) so your AI agents can act with precision tailored to your workflows.
-
Managed MCP Gateway: Publish your APIs safely using a managed MCP gateway that enforces governance.
-
Fine-grain security: Safely expose your Celigo APIs as MCP tools, complete with necessary scopes and authentication.
-
Full compliance: Maintain full governance, safety, and control with comprehensive audit logging of all agent interactions.
-
Isolate by use case: Create any number of MCP Server endpoints, each with a curated tool catalog and environment isolation for specific AI applications.
-
Broadly compatible: Invoke your MCP Server from any MCP-compatible assistant, IDE, or agent framework, regardless of where the agent is built (such as in OpenAI or Claude). Your business logic and APIs become universally accessible to AI systems without custom adapters or lock-in.
-
Simple licensing: All platform editions will include entitlement to create MCP Servers, metered simply as API calls within your existing entitlement.