Data streaming, driven from your IDE - by you and your agents.
Lenses gives engineers and AI agents a single, governed surface for Apache Kafka: a remote MCP server with OAuth 2.1, a VS Code plugin in preview, and an open-source library of Kafka Skills that teach agents how Kafka actually works.
Managing six tools to ship one pipeline?
Building
Agents that write Kafka code without knowing your topic layout, schema evolution, or DLQ patterns ship code that compiles and breaks in production.
Operating
Diagnosing a poison-pill outage or stalled consumer means three consoles, a Slack thread, and a runbook from 2023.
Governing
Static API keys in a config file, no session audit, no way to tell an agent's action from an engineer's.
The full package to make AI work for you
Data streaming through natural language
Connects your AI copilots and agents directly to your Kafka infrastructure & data.
Every agent action governed
Every tool call is attributable - no shared credentials, full audit trail.
Teach your agent how to work with your Kafka
Open source skills that make Claude Code and Cursor Kafka-aware.
Lenses, without leaving your editor
The Lenses experience inside VS Code (and any compatible fork - Cursor, IBM Bob).
Three workflows that used to take half a day
Build - ship a pipeline before standup ends
The agent loads your team's skill file, discovers the environment, checks for topic collisions, creates the topic with a schema, seeds it, deploys the processor, and verifies the output. You watch and approve.
Tools called: list_environments, get_deployment_targets, list_topics, create_topic_with_schema, execute_sql, create_sql_processor.
Add Context...
my-app.py
Operate - debug a poison pill at 3am
An SRE agent with read-only OAuth scopes investigates a stalled consumer: pulls topic metadata, checks consumer group lag, samples bad records, identifies the schema mismatch, drafts a remediation plan. Hand it to a remediator agent with deploy scopes (or do it yourself); approve the SQL processor that filters the bad records into a DLQ.
Investigator and remediator are separate agents with separate scopes. Blast radius limited by design.
Govern - pass an audit without a fire drill
An SRE agent with read-only OAuth scopes investigates a stalled consumer: pulls topic metadata, checks consumer group lag, samples bad records, identifies the schema mismatch, drafts a remediation plan. Hand it to a remediator agent with deploy scopes (or do it yourself); approve the SQL processor that filters the bad records into a DLQ.
Investigator and remediator are separate agents with separate scopes. Blast radius limited by design.
The full package to make AI work for you
Out-of-the-box MCP setups give an agent the same permissions as the engineer running them. That's why 53% of MCP server deployments still rely on static API keys, and only 8.5% use OAuth. The Lenses MCP server flips this.
Up and running in two commands
curl -L https://lenses.io/preview -o docker-compose.yml && ACCEPT_EULA=true docker compose up -d --wait && echo "Lenses.io is running on http://localhost:9991"First tool call triggers the OAuth flow. Approve the scopes on the consent screen. Done.
Lenses 6.2 transforms the interaction with Lenses
