CData
Also known as: CData Software, CData Connect AI, CData AI Gateway, cdata.com
CData gives AI agents governed, live access to enterprise data. Its Connect AI service exposes hundreds of systems such as Salesforce, NetSuite, ServiceNow and Snowflake through one MCP server, where an assistant like Claude, ChatGPT or Copilot can search schemas, query with SQL and insert or update records under each user's own permissions. Roles, toolkits and query logs keep that access governed, and an LLM gateway and context engine are in early access.
CData Software, led by CEO and cofounder Amit Sharma, built its business on drivers and replication tools that connect analytics and integration software to enterprise systems, and it counts more than 10,000 customers. Customer stories name Adobe, GSK, Palantir, Google, Salesforce, UiPath and Anaqua. At Adobe, agents now reach more than 10 systems and serve 20 teams, GSK removed a 90 day wait for IT pipelines, and Anaqua went from evaluation to a company wide rollout of two AI tools in six weeks. CData was named a Challenger in the 2026 Gartner Magic Quadrant for Data Integration Tools.
Connect AI is the product agents use. A single remote MCP server fronts hundreds of sources, so an assistant can find relevant datasets and join hints, list catalogs, schemas, tables and columns, read connector specific guidance, run SELECT queries, call stored procedures and insert or update rows. Every query runs as the requesting user through passthrough identity, so source permissions apply rather than a shared service account.
Administrators can narrow the tools a client sees to read only, publish toolkits that expose chosen tables and operations, model derived views across sources and register third party MCP servers beside CData's own. Setup guides cover Claude, ChatGPT, Microsoft Copilot, Gemini Enterprise, Cursor and Claude Code, and the same data is reachable over REST, OData, ODBC and JDBC.
The AI Gateway adds an LLM gateway and a context engine, both in a limited early access program that accepts organizations by application and runs free for current customers through the end of 2026. The LLM gateway lets administrators register Anthropic, Azure OpenAI, Cohere, Gemini, OpenAI or compatible providers, then route traffic to one provider, through a priority failover chain or against a model allowlist. The context engine keeps company terms and semantic definitions that improve as people correct them. Today, a user context tab lets each person save notes about themselves that the model reads when it answers.
Standard costs $99 a month, or $79 billed annually, for one user and one source, and Growth costs $199, or $159 annually, with premium sources, derived views and toolkits. Business plans add SSO, SCIM 2.0, passthrough identity, pooled tool calls and custom tools on an annual contract. CData holds SOC 2 Type II and ISO/IEC 27001:2022 certifications, offers a HIPAA business associate agreement and keeps no copy of source data.
Query and audit logs record each request and administrative change. Connect AI runs as a managed service, with a Private Cloud Connector for firewalled systems and self hosted drivers for teams that keep connectivity in house. It suits enterprises putting Claude, ChatGPT or Copilot in front of many governed systems at once.
Vendor details
Canonical URL
https://www.cdata.com/
Category
Agent infrastructure
Subcategory
Governed MCP access to enterprise data, with an LLM gateway in early access
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Remote MCP server and REST, OData and SQL interfaces over hundreds of sources such as Salesforce, NetSuite, ServiceNow, Snowflake, Databricks and Workday, with client setup for Claude, ChatGPT, Microsoft Copilot, Gemini, Cursor and Claude Code.
In practice
A finance analyst asks Claude for overdue invoices, and the query runs against NetSuite under her own permissions through the MCP server.
An IT team publishes a read only toolkit so a sales assistant can see Salesforce opportunities but cannot change them.
A platform team audits which user ran each agent query against ServiceNow last week from the query log.
Sources & related URLs
Related / legacy domains
Research sources
Agentic Index coverage score
7.5 / 14 capabilities · 54%
| Integrations & Tool Calling | Full |
|---|---|
|
Through one MCP server an agent can insert rows, update existing rows and run stored procedures in connected systems as well as read them, across hundreds of sources including Salesforce, NetSuite, ServiceNow, Snowflake, Databricks and Workday. Each call runs with the requesting user's identity, so the source system's own permissions decide what goes through, and third party MCP servers can be registered beside CData's connectors. Sourcedocs.cloud.cdata.com/en/api/mcp.mdread 2026-11-20 |
|
| Workflow Orchestration | Not documented |
|
Connect AI answers individual tool calls: search for datasets, list schemas, query, insert, update or run a procedure. The assistant on the other end decides the order of those calls and carries results from one step to the next. Derived views can join several sources into one queryable object, which shapes data rather than sequencing work, and the product runs no workflow, branching logic or coordination between agents. Sourcedocs.cloud.cdata.com/en/api/mcp.mdread 2026-11-20 |
|
| Knowledge Grounding & RAG | Partial |
|
Agents ground answers in live records rather than stale copies: a search tool finds relevant datasets, fields and join hints, connector instructions explain each source's limits, and SQL runs against the system itself at request time. The context engine, which keeps company terms and semantic definitions that improve with corrections, sits in a limited early access program that admits organizations by application. Sourcecdata.com/ai/gateway-early-accessread 2026-11-20 |
|
| Human Oversight & Guardrails | Partial |
|
Administrators decide what an agent may attempt. An ops parameter narrows the advertised tools, for example to read only by leaving out insert, update and procedure calls, toolkits expose only chosen tables and operations, and roles and policies sit on top of source permissions. A user, a connection or the whole account can be revoked from one control plane, though no step asks a person to approve an individual write. Sourcedocs.cloud.cdata.com/en/api/mcp.mdread 2026-11-20 |
|
| Security, Identity & Governance | Full |
|
Connect AI has users, groups, service accounts, workspaces, roles and policies, SAML and OpenID Connect sign in through Okta, Entra ID, Google Workspace, PingFederate or ADFS, and SCIM provisioning on Business plans. Passthrough identity runs each agent query under the individual's own permissions. CData Software holds SOC 2 Type II and ISO/IEC 27001:2022 certifications and offers a HIPAA business associate agreement. Sourcecdata.com/securityread 2026-11-20 |
|
| Observability & Auditability | Full |
|
A query log records every query with its time, type, user, status, connection, query text and destination, with a downloadable detail file kept for seven days. A separate audit log captures administrative events by user, source and target, including changes made over MCP, and SIEM endpoints forward audit events to a security team's own tools. A dashboard summarizes recent queries across the account. Sourcedocs.cloud.cdata.com/en/logs.mdread 2026-11-20 |
|
| Memory & State Persistence | Partial |
|
Each person can save a Markdown note about themselves in a user context tab, and the model reads it when it answers that person. People write and maintain that note by hand. Learned context that grows from user corrections belongs to the context engine, which is limited to organizations admitted to the early access program, and Connect AI keeps no copy of the source data agents query. Sourcedocs.cloud.cdata.com/en/settings/user-context.mdread 2026-11-20 |
|
| Deployment & Data Residency | Partial |
|
Connect AI runs as a managed service, and a Private Cloud Connector reaches databases and applications behind a customer firewall without opening them to the internet. Teams that want connectivity on their own network can run CData drivers through ODBC, JDBC, ADO.NET or Python with no managed service in the loop, though the MCP server, toolkits and governance layer stay in the cloud product, which offers no region choice. Sourcecdata.com/developersread 2026-11-20 |
|
| Prebuilt Agents / Templates / Packs | Partial |
|
Administrators assemble toolkits that expose a chosen set of tables and operations as named MCP tools, and they model derived views that join sources into reusable objects. Universal tools work against any connected source from the first connection. The catalog of hundreds of connectors supplies reach into systems rather than ready made agents or task templates a team adopts as they are. Sourcedocs.cloud.cdata.com/en/api/mcp.mdread 2026-11-20 |
|
| Triggers & Channel Coverage | Partial |
|
Requests arrive from whichever assistant a team connects, with setup guides for Claude, ChatGPT, Microsoft Copilot, Gemini Enterprise, Cursor and Claude Code, and the same data answers REST, OData, ODBC and JDBC calls from applications. Every exchange starts with a client asking. Source changes, schedules and webhooks do not start an agent run here, since change data capture belongs to the separate CData Sync product. Sourcedocs.cloud.cdata.com/enread 2026-11-20 |
|
| Model Flexibility & Routing | Partial |
|
The LLM gateway lets administrators register Anthropic, Azure OpenAI, Cohere, Google Gemini, OpenAI or compatible providers and route traffic to a single provider, through a priority failover chain or against a model allowlist, with budgets set per team or agent. It is limited early access for admitted organizations and pre production software. Outside that program, Connect AI runs no model and serves whichever assistant calls it. Sourcedocs.cloud.cdata.com/en/llm-gateway/llm-gateway.mdread 2026-11-20 |
|
| APIs / SDKs / MCP Extensibility | Full |
|
The remote MCP server has a fixed endpoint, OAuth or personal access token sign in and an enumerated tool set, from search and schema discovery to insert and update, and a tools list request shows what each account can call. Connect AI also exposes a REST API, OData and SQL through ODBC and JDBC, and a management MCP server administers the account itself. Sourcedocs.cloud.cdata.com/en/api/mcp.mdread 2026-11-20 |
|
| Testing, Debugging & Optimization | Not documented |
|
CData Labs runs its own accuracy and efficiency benchmarks comparing Connect AI with other MCP approaches, which describe the product rather than a customer's agent. Query logs and a downloadable detail file help troubleshoot a failed call after it happens. The service has no test harness, scored cases or release gate for checking how an agent behaves on a team's data before it goes live. Sourcedocs.cloud.cdata.com/en/logs.mdread 2026-11-20 |
|
| Browser / Computer-use | Not documented |
|
Every source is reached through an API or database connector that CData maintains, which is the point of the product: agents query and write structured data without screen scraping. Results come back as CSV through tools rather than through a page an agent reads. No browser, desktop or remote computer session is part of Connect AI, and applications without an API stay out of reach. Sourcedocs.cloud.cdata.com/en/api/mcp.mdread 2026-11-20 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Pricing
$79 a month billed annually, or $99 monthly; Growth $159 or $199
Users and data sources by plan
Included quota
Standard includes one user and one standard data source. Growth adds Growth and Premium sources, derived views and toolkits.
What is public
CData sells Connect AI in three plans. Standard costs $99 a month, or $79 a month billed annually, for one user and one standard source. Growth costs $199, or $159 billed annually, and opens Growth and Premium sources, derived views and toolkits.
Billing mechanics
Annual billing takes 20% off the monthly price. Extra users and sources are bought as add ons. Business plans run on an annual contract with a custom number of sources and pooled tool calls.
Overage / add-ons
Extra users and data sources are sold as add ons on Standard and Growth.
Sales call required
Mixed (some tiers require a call)
Free / trial
Free trial
Lowest paid plan
Standard
Commercial notes
Business plans add premium support, a dedicated customer success manager, passthrough identity, SCIM 2.0, enterprise SSO, regulated industry compliance and custom tools, with volume discounts. The LLM gateway and context engine are in an early access program that admits organizations by application and is free for current customers through December 31, 2026, with general availability pricing to follow.
Missing data
Business plan prices are set in a contract after a sales conversation.
Related vendors
- AgentOps — Agent observability and debugging platform: open source SDKs trace…
- AgentPhone — AgentPhone gives the AI agents a team builds their own US or…
- Agno — Python agent framework and AgentOS runtime (formerly Phidata) for…
- Airbyte — Airbyte moves data from operational systems into warehouses and…
- AIsa — Resource and payment gateway for AI agents, with one key to 110+…
- Albato — Albato is a no code integration platform that sells an embedded,…