Couchbase AI Data Plane

What it does

The Couchbase AI Data Plane takes in text and structured or unstructured data stored in Couchbase clusters and produces embeddings and model inference outputs such as chats and completions. Its Model Service deploys LLMs and embedding models either as serverless models billed per token or as provisioned models running on dedicated GPUs billed by the hour. The Data Processing Service and Workflows vectorize data for use with other AI Data Plane features.

Developers use it to build agentic applications: Agent Memory provides a persistent memory layer that recalls context across sessions, Agent Catalog manages tools and prompts for custom agents, and the MCP Server lets LLMs interact directly with data in Couchbase clusters through tools. AI Functions let SQL++ queries summarize, classify, or detect sentiment in text. Management and Model Service REST APIs handle deployment and inference without special SDKs.

This suits teams already using Couchbase Server Enterprise Edition or Capella who want model hosting and agent data services in the same environment. Access requires a Capella account or self-managed deployment; the Management API is unavailable on free-tier operational clusters. Key limits include a default of 100 requests per minute per management API key, a default of 1,000 model calls per minute, and guardrails available only on provisioned LLMs.

What it can do

• Deploy serverless LLMs and embedding models billed per token from Amazon Bedrock
• Deploy provisioned LLMs and embedding models on dedicated NVIDIA NIM GPUs
• Persistent Agent Memory layer for recalling context across user sessions
• Agent Catalog for managing tools and prompts for custom agents
• AI Functions for summarizing, classifying, and detecting sentiment within SQL++ queries
• Management API and Model Service API for deployment and inference over REST

Works with

• OpenAI
• Amazon Bedrock
• Couchbase Capella
• Couchbase Server Enterprise Edition
• AWS PrivateLink

Before you choose

• Serverless models cannot be named or resized, and changing a deployment such as caching requires deleting and redeploying it; guardrails and jailbreak detection are provisioned-LLM only.
• Workflows can use only provisioned embedding models, and the Management API is not available on free-tier operational clusters.

Listed under

• Coding
• MCP and agent tools

Get started

Get started by creating a Management API key and making a call; no special SDKs are required.

1. Create a Capella account or prepare a self-managed Couchbase Server Enterprise Edition deployment.
2. Create a Management API key with the appropriate Capella roles and an optional IP allowlist, then pass it as a Bearer token.
3. Deploy a serverless or provisioned LLM or embedding model, then use its Model Service API key and connection string to send inference requests.

Serverless models bill per input and output token in Capella credits; provisioned models bill by the clock hour based on compute size and whether the model is on. Verify rates and enterprise support licensing cost with Couchbase before purchase.

We couldn’t find published prices for Couchbase AI Data Plane.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Couchbase AI Data Plane | Couchbase Docs ↗docs.couchbase.com
• Deploy Models with the AI Data Plane Model Service | Couchbase Docs ↗docs.couchbase.com
• Manage Deployments with the Couchbase AI Data Plane APIs | Couchbase Docs ↗docs.couchbase.com

Company website

docs.couchbase.com: product, pricing and policy pages

Web archive

When the site first appeared

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

manual_review

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.