Teradata
Also known as: teradata.com, Teradata Corporation, TDC, Teradata Tera, Teradata Enterprise AgentStack, Teradata AgentBuilder, Teradata Enterprise Vector Store, Teradata MCP Server, Teradata Autonomous Knowledge Platform
Teradata is a data platform company whose agent surface today is the data layer agents run on: Enterprise Vector Store, with multimodal ingestion and hybrid search inside the database, and an open source MCP server that gives any agent governed tools over Teradata data. Its own agents, the Tera workspace and Enterprise AgentStack with ready SQL, data science and monitoring agents, are in early access. It runs in the cloud, on premises or in sovereign setups, priced by quote in Teradata Units.
Teradata, the NYSE listed data platform company based in San Diego, is reshaping its warehouse into a knowledge platform for AI agents. It counts all ten of the top 10 banks and seven of the top 10 airlines, telecom companies and healthcare providers among its customers, with American Airlines, Elevance Health, HSBC, Vodafone and Standard Chartered among the names.
Its shipped agent surface today is the data layer agents run against: Enterprise Vector Store, generally available with multimodal and agentic functions since April 2026, and an open source MCP server that lets any agent query, profile and govern Teradata data. Its own agents, the Tera workspace and the Enterprise AgentStack toolkit, are offered through early access. Teradata has built a conversational analytics layer for a large European bank and network fraud detection for an Asia Pacific bank.
Enterprise Vector Store runs inside the database. It ingests documents, PDFs, images and audio through Unstructured in more than 64 file formats, generates embeddings, indexes them beside structured tables and serves hybrid semantic, keyword and metadata search, with LangChain integration for retrieval and agent workflows, at a scale of billions of vectors and more than 1,000 concurrent queries.
The MCP server, released under the MIT license and run with uv, Docker or as an HTTP service, gives agents tools for querying and exploring tables, data quality, plotting, retrieval and vector search, database administration, security and backup, with semantic layers defined in YAML, row limits and row level access views.
Through early access, Enterprise AgentStack adds AgentBuilder, with no code building through Karini AI and pro code through Flowise, CrewAI and LangGraph, ready SQL, data science, monitoring and customer lifetime value agents and a bring your own model option; AgentEngine, a Kubernetes or Docker runtime with shared memory for single and multi agent systems; AgentOps for monitoring, policy enforcement and human in the loop control; and Enterprise MCP, a curated tool set. Tera, the agentic coworker, works from the Console, AI Studio and notebooks with a context engine, harness and skills for exploring, analyzing, building and optimizing, and Tera Agents for sizing, telemetry, FinOps, tuning and compute aim to run the platform on their own.
Teradata holds SOC 1 and SOC 2 Type II reports and ISO 27001 and 27017 certification. It is also audited for payment card security and for health data under HITRUST and HIPAA. It runs on Teradata Cloud, in customer managed clouds, on premises through Teradata Factory with Dell and NVIDIA hardware, and in sovereign setups. Pricing is quoted in Teradata Units under a fixed plus flex model, with a calculator and proof of value programs. It fits large enterprises already running Teradata that want agents to work on governed data where it sits; teams wanting a finished analyst agent today will find most of it still in early access.
Vendor details
Canonical URL
https://www.teradata.com
Category
Agent infrastructure
Subcategory
Data platform whose shipped agent surface is Enterprise Vector Store and an open source MCP server for agents, with the Tera agentic coworker and Enterprise AgentStack (AgentBuilder, AgentEngine, AgentOps, Enterprise MCP) in early access
Funding status
Public company (NYSE: TDC), headquartered in San Diego.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Open source Teradata MCP Server for any MCP client; LangChain integration for Enterprise Vector Store; Unstructured for multimodal ingestion; Karini AI, Flowise, CrewAI and LangGraph in AgentBuilder (early access); Pinecone partnership; NVIDIA NeMo Retriever; Fabric (formerly QueryGrid) across data systems; Python and SQL interfaces.
In practice
A bank's data team connects its internal analyst agent to Teradata through the open source MCP server, so the agent profiles tables, checks data quality and runs queries under the service account's row level access views instead of copying data out.
An insurer loads claim documents, adjuster photos and call recordings into Enterprise Vector Store beside its policy tables, and its claims agent runs one hybrid search across both to explain why a claim was flagged.
A retailer in the AgentStack early access program builds a customer lifetime value agent in AgentBuilder from the ready template, deploys it on AgentEngine in its own Kubernetes cluster and watches it in AgentOps.
Sources & related URLs
Related / legacy domains
Research sources
Agentic Index coverage score
7.5 / 14 capabilities · 54%
| Integrations & Tool Calling | Partial |
|---|---|
|
The open source MCP server connects any agent to Teradata data with tools for querying, profiling, data quality, retrieval, administration, security and backup, and Fabric links Teradata to other data systems. AgentBuilder, in early access, adds Flowise, CrewAI, LangGraph, Karini AI and cloud provider tools, but no shipped agent acts in business systems outside the data platform. Sourcegithub.com/teradata/teradata-mcp-serverread 2026-11-19 |
|
| Workflow Orchestration | Partial |
|
AgentBuilder, in early access, builds single and multi agent workflows from start, condition and agent nodes, as in a credit risk profiling example, and AgentEngine runs them on Kubernetes or Docker with shared execution workflows. Both are gated, so orchestration a customer can run today comes from the open frameworks it brings, with Teradata supplying the data tools underneath. Sourceteradata.com/insights/ai-and-machine-learning/agentic-ai-stack-with-teradata-agentbuilderread 2026-11-19 |
|
| Knowledge Grounding & RAG | Partial |
|
Enterprise Vector Store ingests documents, PDFs, images and audio in more than 64 formats, generates embeddings and indexes them inside the database next to structured tables, serving hybrid semantic, keyword and metadata search with lineage kept for citations. It is the means for a team to build retrieval for its agents rather than a maintained knowledge base the agents draw on by themselves, and Tera's context engine with governed definitions is in early access. Sourceteradata.com/press-releases/2026/teradata-enables-ai-agentsread 2026-11-19 |
|
| Human Oversight & Guardrails | Partial |
|
AgentOps, in early access, enforces enterprise policies and compliance checks on agents and adds human in the loop controls, and Tera Agents act within policy governed autonomy in a secure harness. On the shipped MCP server, the guardrails are database ones: row limits on results, row level access views and the permissions of the account an agent connects with. Sourceteradata.com/platform/agent-stackread 2026-11-19 |
|
| Security, Identity & Governance | Full |
|
Teradata holds SOC 1 and SOC 2 Type II reports and ISO 27001 and 27017 certification. It is also audited for payment card security and for health data under HITRUST and HIPAA, and meets Australian and Texas government security programs. Its cloud uses role based access, multi factor authentication for remote administration and FIPS 140 2 validated encryption, and agents reach data through governed tool access. Sourceteradata.com/trust-security-center/data-complianceread 2026-11-19 |
|
| Observability & Auditability | Partial |
|
AgentOps, in early access, gives one place to discover, monitor and manage the lifecycle of agents across an organization, and Enterprise MCP adds monitoring of tool use. ModelOps, already shipped, monitors machine learning models in production for drift and performance, but traces of an agent's reasoning, tools and actions arrive only with the gated stack. Sourceteradata.com/platform/ai-studioread 2026-11-19 |
|
| Memory & State Persistence | Partial |
|
AgentEngine, in early access, gives agents persistent memory over structured and unstructured data and lets agents in one runtime share memory, data structures and workflows. Shipped today, the database and vector store can hold whatever state a team's own agents write to them, but no generally available Teradata agent keeps memory across sessions. Sourceteradata.com/platform/agent-stackread 2026-11-19 |
|
| Deployment & Data Residency | Full |
|
The platform runs on Teradata Cloud, in customer managed clouds, on premises through Teradata Factory built on Dell PowerEdge servers and NVIDIA AI infrastructure, and in sovereign deployments for strict residency rules. The MCP server runs on the customer's own machines through uv or Docker, and AgentEngine deploys to Kubernetes in the cloud or on premises. Sourceteradata.com/platformread 2026-11-19 |
|
| Prebuilt Agents / Templates / Packs | Partial |
|
AgentBuilder includes ready agents for SQL generation, data science pipelines, database monitoring and customer lifetime value, usable as templates, and Tera Agents cover sizing, telemetry, FinOps, tuning and compute. All of them are in early access, and the shipped MCP server offers groups of tools rather than finished agents a team can switch on for a task. Sourceteradata.com/platform/agent-stackread 2026-11-19 |
|
| Triggers & Channel Coverage | Partial |
|
Tera Agents are built to run continuously, sizing compute to workload and service levels and watching telemetry and spend, and AgentBuilder flows start from chat or form inputs. Both sit in early access; the shipped MCP server and vector store answer the agent that calls them and start nothing on schedules, alerts or business events of their own accord. Sourceteradata.com/press-releases/2026/introducing-the-autonomous-knowledge-platformread 2026-11-19 |
|
| Model Flexibility & Routing | Partial |
|
AgentBuilder supports bring your own model, with language models from cloud providers, open model APIs or on premises NVIDIA deployments, and Tera offers a choice of model as well. Both are early access, so the shipped surface leaves the model to whatever agent the customer brings to the MCP server, and no routing between models by cost or task comes with either product. Sourceteradata.com/platform/agent-stackread 2026-11-19 |
|
| APIs / SDKs / MCP Extensibility | Full |
|
Teradata's MCP server is open source under the MIT license and runs locally, in Docker or as an HTTP service, with tool groups for query, quality, plotting, retrieval, vector store, administration, security and backup, plus semantic layers defined in YAML. Enterprise Vector Store integrates with LangChain, and Python and SQL interfaces open the rest of the platform to a team's own agents. Sourcegithub.com/teradata/teradata-mcp-serverread 2026-11-19 |
|
| Testing, Debugging & Optimization | Not documented |
|
AI Studio, not yet generally available, includes testing and an evaluation environment among its safeguards, without test sets, scoring or comparison of agent versions a team can work with today. Nothing shipped lets a team test an agent against expected answers before release, and Tera's own benchmark compares token use with another coding agent rather than giving customers a test tool. Sourceteradata.com/platform/ai-studioread 2026-11-19 |
|
| Browser / Computer-use | Not documented |
|
Agents reach Teradata through SQL, MCP tools and framework connectors such as LangChain, and the company's own agents act inside its data platform. None of the shipped or early access products drives a browser or desktop session to work through consoles or web applications, so work in tools without a data or tool connection stays outside their reach. Sourceteradata.com/platform/teraread 2026-11-19 |
|
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
Quote only; Teradata Units under fixed plus flex consumption; open source MCP server free
Teradata Units, one currency across compute, storage, software and AI, with a committed fixed baseline plus flex consumption
What is public
Teradata prices by quote in Teradata Units, one currency across compute, storage, software and AI, with a committed fixed baseline and flex capacity paid as used. A calculator gives estimates, and the community MCP server is free under the MIT license.
Billing mechanics
Package upgrades take effect at any time and downgrades at the next renewal, across three packages and two storage options.
Cost watchouts
AgentStack features arriving in 2026 carry additional costs on top of the platform.
Variable cost rationale
Use above the fixed commitment is billed as flex consumption in Teradata Units, so agent workloads that query heavily raise the bill.
Additional watchouts
Agent tooling in AgentStack is priced on top of the platform as it ships.
Overage / add-ons
Flex capacity is paid as consumed above the fixed commitment.
Sales call required
Yes, required for paid access
Free / trial
Proof of value programs; the community MCP server is free and open source under MIT
Commercial notes
Proof of value programs let organizations evaluate the platform before committing.
Key ambiguities
The dollar value of a Teradata Unit is not public.
Cancellation / refund
Package downgrades take effect at the next contract renewal.
Missing data
The price of a Teradata Unit, the package prices and AgentStack prices are not public.
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,…
Alternatives to Teradata
Other products that do the work Teradata does, and what each one is.