AGENT INFRASTRUCTURE

Real-Time Data for AI Agents

AI agents need fresh data, not stale snapshots. Streamkap streams database changes to your agents in real time through MCP, CLI, API, and native Streaming Agents, so every decision is based on what is happening now.

Why Agents Need Real-Time Data

Batch ETL = Stale Context

Agents querying warehouses refreshed hourly make decisions on outdated data. A customer could churn, an order could fail, and the agent would not know for hours.

Warehouse Latency is Too High

Data warehouses return results in 1-10 seconds. Agents need sub-100ms context lookups to maintain conversational speed and real-time responsiveness.

Streaming Keeps Agents Current

Streamkap captures every database change as it happens and delivers it in sub-second latency. Agents always operate on the latest state, not yesterday's snapshot.

Connect Agents Your Way

Three interfaces for agent connectivity, from conversational AI tools to custom integrations.

Available Now

MCP Server

Connect Claude, Cursor, VS Code Copilot, and any MCP-compatible AI tool to your streaming pipelines. Ask questions about pipeline health, connector status, and throughput in natural language.

  • Hosted at mcp.streamkap.com, no local install needed
  • Monitor pipelines, inspect connectors, query metrics
  • Works with Claude Desktop, Claude Code, Cursor, Windsurf
  • Automatic credential management with JWT tokens
MCP Server Docs
Available Now

CLI

Manage pipelines from the terminal. Script deployments, automate health checks, and integrate streaming infrastructure into CI/CD workflows. Outputs JSON for agent consumption.

  • Install via npm: npm install -g @streamkap/tools
  • 15+ command groups: pipelines, sources, destinations, transforms
  • JSON output mode for scripting and agent integration
  • Environment variable auth for CI/CD automation
CLI Docs
Available Now

REST API

Direct programmatic access to everything in Streamkap. Build custom agent integrations, automate pipeline management, and query metrics from any language or framework.

  • Full OpenAPI spec for code generation
  • Manage pipelines, connectors, transforms, and schemas
  • Query metrics, logs, and consumer group state
  • Scoped API tokens for fine-grained access control
API Docs

Coming Soon

In Development

Streaming Agents

Run AI agents directly inside Streamkap's processing engine. Workflow Agents for deterministic multi-step pipelines and ReAct Agents for autonomous reasoning, both with exactly-once guarantees and built-in state management.

  • Workflow Agents: deterministic, multi-step pipelines with branching and retries
  • ReAct Agents: autonomous reasoning with tool calls and state persistence
  • Exactly-once processing guarantees with checkpoint recovery
  • Native access to all streaming data, no external API calls needed
In Development

Agent Storage

A purpose-built storage layer optimized for agent workloads. Sub-100ms queries on streaming data, designed for the read patterns agents actually use: point lookups, time-range scans, and context assembly.

  • Sub-100ms query latency on live streaming data
  • Built-in MCP and API access for agents
  • Automatic materialization from streaming pipelines
  • Up to 90% cost reduction vs warehouse queries for agent workloads
In Development

Decision Governance

Every autonomous agent decision needs an audit trail. Streamkap's governance layer records what data triggered each decision, what action was taken, and what the outcome was, giving enterprises the accountability they need before deploying agents at scale.

  • Decision Traces: Full chain from data event to agent action to outcome, stored as an immutable log
  • Policy Gates: Configurable rules that intercept agent actions before execution, like approval workflows for high-value decisions
  • Replay & Debug: Re-run any past decision with the same inputs to understand exactly what happened and why
  • Access Scoping: Control which agents can read which data streams and which can trigger actions
  • Anomaly Detection: Monitor agent decision patterns over time and flag deviations from expected behavior
  • Compliance Export: Export decision traces in audit-ready formats for SOC 2, EU AI Act, HIPAA, and financial regulations

What Agents Can Do with Streamkap

Monitor Pipeline Health

AI assistants check connector status, track replication lag, and flag anomalies before they become incidents. Ask Claude: "Are any pipelines behind?"

Real-Time Fraud Detection

Streaming Agents process transaction events continuously, applying ML models and business rules to flag suspicious activity within milliseconds.

Inline Inference

Run ML models and LLM calls directly on streaming events as a transform step. Sentiment analysis, classification, entity extraction, and scoring happen inline before data reaches its destination.

Customer 360 Context

Stream events from CRM, billing, and product databases into a unified context store. Agents always have the full picture when interacting with customers.

Automated Incident Triage

When pipeline metrics degrade, agents automatically inspect logs, identify root causes, and execute remediation steps through the MCP Server, CLI, or API.

Event-Driven Automation

Trigger agent workflows from database changes. A new order, a status update, or a config change becomes an event that agents act on immediately.

The Agent Data Stack

Streamkap provides the complete streaming layer between your databases and your agents.

Sources
PostgreSQL, MySQL, MongoDB, DynamoDB
Streamkap
Streaming + Processing + Storage + Governance
Connect
MCP / CLI / API / Streaming Agents
Agents
Claude, GPT, Custom Agents, Streaming Agents

Developer Tools

Terraform Provider

Manage Streamkap resources as infrastructure-as-code alongside your application stack.

Terraform Docs

API Tokens

Generate scoped credentials for agents, CI/CD pipelines, and automated workflows.

Token Docs

OpenAPI Spec

Generate client libraries in any language from the full OpenAPI specification.

API Reference

Connect Your Agents to Real-Time Data

Start streaming events to your agents in minutes. MCP Server, CLI, and API are available today.