- Tool type
- Event-driven AI agents on Flink and Kafka
- Made for
- Data and AI engineers building event-driven automation
- Runs on
- Confluent Cloud, managed Apache Flink and Kafka
- Agents execute as Flink jobs within Confluent Cloud; the web console and CLI are used to manage them.
What it does
Streaming Agents are AI agents that run as Apache Flink jobs inside Confluent Cloud. They take in live event streams from Kafka topics and enrichment sources such as internal systems, vector or text search, and tool calls via Model Context Protocol. They produce actions, such as updated records, notifications, or escalations, along with a replayable event log of what the agent did.
People use Streaming Agents for workflows that need to react as events happen, such as real-time product personalization, anomaly detection with root-cause assessment, and automated competitive price matching. The distinguishing feature is that agents sit in the same pipeline as stream joins and filters, so engineers can build AI steps with familiar Flink APIs and SQL rather than a separate agent framework.
This suits engineers and data teams already using Confluent Cloud, Kafka, or Flink who want event-driven automation with LLMs. Access requires a Confluent Cloud account, with a sign-up offer of $400 credit for the first 30 days. Limitations include a dependency on Confluent's managed stack, and the provider’s documentation do not confirm the model list, regions, or long-term pricing.
What it can do
- Embed LLM and embedding model calls directly in Flink stream processing pipelines
- Tool calling through Model Context Protocol (MCP) and A2A
- Data enrichment from external systems via External Tables and vector, text, and key search
- Event-driven multi-agent patterns such as Orchestrator-Worker and Hierarchical
- Replay of Kafka's immutable log to rewind, debug, test, and audit agent behavior
- Secure connections, RBAC, and dark mode testing for governed agent flows
Works with
- Model Context Protocol (MCP)
- A2A
- External Tables & Search (vector search, text search, key search)
Before you choose
- Streaming Agents depend on Confluent Cloud's managed Flink and Kafka services; the provider’s documentation do not describe a standalone deployment option.
- The $400 credit is described as a 30-day offer; ongoing costs after the credit were not confirmed in the provider’s documentation.
- Model choice, cloud region, and tier availability were not confirmed on the provider’s documentation and should be checked in the docs before planning.
Listed under
Get started
Sign-up includes $400 to spend during the first 30 days, per the product page.
- Create a Confluent Cloud account at the Get started page, or sign in with Google or GitHub.
- Review the Streaming Agents docs and the quickstart repository linked from the product page.
- Set up a Kafka cluster and Flink compute on Confluent Cloud, then connect your model endpoints and tools.
Plans and prices
The product page offers $400 to spend during the first 30 days. Usage-based rates were not found on the provider’s documentation; the pricing page linked from Confluent is not in the provider’s documentation.
We couldn’t find published prices for Confluent Streaming Agents. See their pricing page ↗
30-day free trial.
Sources for this overview
Product pages, documentation, and pricing checked 30 Oct 2026. Each source supports part of the overview; availability and prices can change.
- Streaming Agents on Confluent Cloud | Confluent ↗www.confluent.io
- Confluent Documentation | Confluent Documentation ↗docs.confluent.io
Ratings and reviews
No ratings for Confluent Streaming Agents yet.
What’s changed
Free accountSignals
Free accountData and security
Free accountCompare with similar tools
Free accountWhere this comes from
How we research- Company website
- confluent.io: product, pricing and policy pages
- Social accounts
- Accounts the company links, and its posts on X
- Human review
- Checked by a person
- knowledge_base
- manual_review
Last checked 30 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.