Monarch vs CData Connect AI for Enterprise Agents

CData Connect AI gives agents access to business data and supported actions. Monarch helps agents carry out complete business processes across applications, including the steps existing connections cannot reach.

Monarch is an AI agent integration platform for connecting the steps of a business process across modern, legacy, and internal applications. The aim is to let an agent finish work that still depends on people moving between those systems.

CData starts with connections to business data and the queries and actions those connections support. It can write data as well as read it.

Monarch’s Product Graph maps how applications behave: the actions they allow, what those actions require, and how one step affects another. Releasing a held order might require a credit check in an old finance system, a stock reservation, and a manager’s approval before the order system will accept the change.

That is a different technical starting point from exposing connected data through a common interface. Monarch discovers and maintains the application knowledge needed to complete the process. If CData already exposes all the actions your job needs, its connectivity may be enough.

What each platform is for

CData access and actions, API Connector, Monarch Product Graph.

Compare Connect AI with the job you need done

This comparison focuses on CData Connect AI, the company's managed platform for agent connectivity. Its interfaces include MCP, SQL-oriented access, and REST/OData, so the same connectivity can serve agents and other applications. CData Connect AI documentation

The CData product in a proposal matters. A driver's database interface, a managed MCP service, and an integration product can sit in different parts of your architecture. Confirm which product supplies the actual operation before comparing implementation effort or deployment requirements.

CData supports writes as well as queries

CData Connect AI can join information across connected sources and supports write, update, and delete operations. The available operations depend on the source system's API capabilities. That makes the specific connector and operation more useful evaluation criteria than a company-wide read-versus-write label. CData data access and actions

Its custom-tool guide includes a concrete write example: inserting a support-ticket record into a connected Google Sheet through a parameterised SQL tool. That demonstrates a configured write path. Your production application needs its own supported operation, permissions, and validation. Custom MCP tool example

Take the held order. Updating a field called Status might perform the required transition. It might also leave out a credit check, inventory reservation, or approval that normally happens when an operator releases the order.

Have the application owner demonstrate the intended operation and the evidence it leaves behind. This is a test for both products. A field update is useful when it produces the business result you approved.

Curated tools can make known operations reusable

Connect AI's toolkits let administrators choose the actions exposed to an agent. They can include universal tools, source-specific tools, and custom SQL operations with defined parameters. The source-tool documentation includes creating, updating, and deleting Jira issues. CData toolkits

Keep the business meaning visible in that definition. An operation that retrieves orders awaiting credit review should establish the relevant business unit and review state. The agent should not have to guess whether a blank approval field means pending, rejected, or exempt.

A proprietary API may already be enough

CData's API Connector can connect to proprietary REST endpoints. Its configuration covers authentication, headers, pagination, and the tables exposed to consumers. A missing entry in the packaged source catalogue does not automatically rule it out. CData API Connector

Monarch's Product Graph addresses the application knowledge behind the operation. It discovers public APIs, private web capabilities, and UI paths available to the connected account, then represents the operations captured and verified through discovery.

That becomes useful when an operator knows how to complete the job but the implementation team lacks a tested way to automate it. A release screen may depend on several records, a prior approval, and a private application operation that isn't described in the public documentation.

The application map gives the workflow that operational context. Your team still supplies the authoritative records and business policy. Repeatable steps can be captured in deterministic code once the path is established.

Test the work saved in discovering, verifying, and maintaining that path. If a configured CData tool already performs it correctly, rebuilding it needs a concrete justification.

Test the identity behind each operation

CData Connect AI supports shared authentication and per-user authentication for supported sources. Its OData API requires shared authentication. It also has separate permissions for selecting, inserting, updating, deleting, and executing procedures. These are useful controls, but the selected authentication pattern changes whose source permissions apply. CData access control

For the order workflow, establish whether the agent acts through a dedicated service account or an individual user's connection. Then test an order that identity can read but must not release.

In a Monarch evaluation, use the intended account during discovery and execution. A pilot demonstrated with an administrator account gives you little evidence about the permissions your operations team will approve.

For both designs, approval of a particular release needs to be part of the workflow. Permission to update orders doesn't establish that this order has passed the required review.

Connect the investigation to the work that follows

An organisation may already have useful data connections while the operations that resolve an exception still depend on people. Monarch expands discovery into those accessible application steps and orchestrates agentic workflows across the verified map.

For an order on hold, that means connecting the evidence to an approved release and the final case update. Other workflows can reuse the customer, order, and invoice operations, with their own prerequisites and controls. Monarch maintains the application paths those workflows depend on as products change.

CData can continue supplying connections and tools that already meet the requirement. Design the handoff around the work each platform owns. A reusable credit-review pattern is useful only after it reflects your legal entities, approval policy, and conditions for releasing an order.

Follow a held order through the whole decision

Consider a hypothetical distributor investigating an order on hold. The customer record sits in CRM, outstanding invoices are in finance, and an internal fulfilment application holds the release decision.

The agent's first task is to prepare the case for a credit reviewer. It needs the correct customer and legal entity, the order reference, and the relevant invoice evidence. A payment recorded against another entity should remain an exception for review.

With CData Connect AI, start by testing the connections and queries that assemble that evidence. Then identify the release operation: an available source tool, a custom SQL operation, or a separately implemented route into the fulfilment application.

With Monarch, scope discovery to the same records and required actions. Prove that the workflow can retrieve the evidence, present the review, and invoke the authorised release operation under the intended account.

Give the reviewer a case they can inspect, including unresolved mismatches. Define what their approval permits. If the order value or customer reference changes afterwards, the workflow needs to detect that change and apply the agreed revalidation rule before acting.

Completion should include the release state in fulfilment and its reference recorded against the case. If the release succeeds but the case update fails, retain evidence of the release and report the unfinished step.

Now repeat the test with an expired credential, a rejected approval, and a response that arrives too late. Ask the implementation team to show how it establishes the order's actual state before retrying. Count the operator's recovery work alongside the successful runs.

Match the evaluation to the work that remains

For an established workflow, Monarch captures repeatable steps in code and maintains the application map those steps rely on. That reduces repeated discovery during execution and supports focused repair as applications change. Test the order release after changing a required application field, including the time to repair and verify the result.

Keep working integrations in the comparison. The cost of extending an existing CData deployment can be very different from building the same access again elsewhere.

Trace the deployment for your Australian or US team

CData's Connect Gateway runs in a private network and creates a reverse tunnel to Connect AI's cloud services. It supports Docker and Kubernetes deployment and lets supported sources remain behind a firewall. The gateway's location therefore answers one part of the deployment question; the connected cloud service and consuming agent also need to be assessed. Connect Gateway

For an Australian deployment, ask where the selected Connect AI service processes requests, where evidence is retained, and which regions the agent's model uses. Make the same checks for a US deployment. Specify the required service locations in the proposal rather than inferring them from where the source database runs.

Monarch offers hosted AWS deployment in the US and Australia, with customer-cloud hosting as an Enterprise add-on. Agree the model-processing, logging, and support-access requirements during the security discussion.

For either design, have the team trace one case from source access through the final write and retained evidence. That makes the responsibilities clear enough for your application and security owners to review together.

Price the configured workflow

CData publishes Standard, Growth, and Business plans. Its pricing distinguishes source tiers, users, and capabilities, with Business plans handled through a sales conversation. Ask for the exact sources, tool configuration, and identity controls used in your proposed workflow to be included in the quote. CData Connect AI pricing

Monarch uses platform and per-action pricing. Include reads, writes, retries, and deployment requirements when scoping the pilot.

Compare the total cost of accepted cases, including the agent's model use and the people who review exceptions. Track the initial implementation separately from ongoing operation, and include the work to verify a changed application.

If CData's connections and curated tools already complete the job, that is a sound route to evaluate. Where the remaining work depends on poorly understood application behaviour across systems, give Monarch that specific problem to prove.

Bring one held-order case, the evidence your reviewer needs, and the release step your team still handles manually. We'll use that case to scope the access, approvals, and completion checks for a Monarch pilot.