Agents

Autonomous agents that compound knowledge over time.

Riven Agents are production-grade autonomous systems with tiered persistent memory — short-term context, long-term pgvector storage, and shared workspace memory. Full MCP tool-use gives agents access to any API, database, or external system.

3-Tier
Memory Model
MCP
Tool Protocol
N×N
Multi-Agent
Loading demo…

What's included

Everything you need, nothing you don't.

Tiered Memory

Short-term scratchpad, long-term pgvector store, and shared workspace memory. Agents remember context across sessions and between runs.

MCP Tool Use

Plug any MCP-compatible tool into your agent — APIs, databases, file systems, browsers, and custom tools.

Multi-Agent Orchestration

Spawn sub-agents, fan out parallel tasks, and aggregate results. Orchestrator handles routing, retries, and failure recovery.

Safe & Observable

Every tool call is logged, every decision is traced. Set guardrails, approval gates, and spend limits at the agent level.

Full capability list

Three-tier memory (short/long/shared)
pgvector semantic memory store
MCP-native tool execution
Multi-agent fan-out orchestration
Retry and fallback strategies
Spend limits & rate controls
Human-in-the-loop approval gates
Full decision trace & replay
Webhook triggers & scheduled runs
REST API for agent management

Early access

Riven is in beta — pricing opens as we leave beta. Request access and we'll reach out within a few days.

Cookie Preferences

We use essential cookies to operate the site. Optional cookies help us improve your experience. Cookie Policy