LoginSign Up
← Back to all sparks
R

Rivet

DEVOPS
Velocity7.5

Rivet is shipping the complete agentic backend stack: actors, durable streams, BYOC, MCP integration, and a V8 app runtime in one month.

actor-platformagentic-backendbyoc-enterprisedynamic-appsmcp-integration
◆Current state
Rivet is in an intense release sprint across every layer of its platform. In roughly 25 days: JWT authentication for actors with direct client connections, OpenTelemetry tracing for actors and workflows, BYOC enterprise deployment into customer AWS/GCP VPCs, MCP integration for AI coding tools, Durable Streams on actor infrastructure, and Dynamic Apps (a V8 isolate runtime for deploying user-generated AI applications). Older entries reveal a zero-disk S3-tiered SQLite engine and an agentOS execution API for JavaScript and Python. The product is moving faster than almost any comparable infrastructure platform.
◆Where it's heading
Rivet is building toward a complete backend platform for AI applications — one where persistent stateful actors, durable messaging, serverless AI app deployment, enterprise security, and native AI tool integration all run under one abstraction. Dynamic Apps plus Rivet Actors could become a unified PaaS where both AI-generated user-facing apps and their backend state live on Rivet infrastructure. The BYOC offering opens enterprise accounts that require data residency, and the MCP integration ensures Rivet is visible from within AI coding environments where developers make infrastructure decisions.
◆Prediction
Dynamic Apps will expand to Python (matching agentOS's recent JavaScript + Python execution API addition), and Rivet will consolidate billing across actor-hours and Dynamic App compute into a single predictable spend model. BYOC pricing will be announced publicly as the enterprise motion becomes more defined.

◆Recent moves

  1. 3d ago

    Introducing JWT Authentication for Rivet Actors

    JWT authentication for Rivet Actors adds short-lived, actor-scoped tokens with direct integrations for Auth0, Clerk, and Better Auth — enabling client-to-actor connections without routing credentials through a server intermediary. This extends the actor model to user-facing scenarios where individual users need isolated, authenticated state.

  2. 4d ago

    Introducing OpenTelemetry for Rivet Actors and Workflows

    OpenTelemetry for Rivet Actors and Workflows adds first-class distributed tracing with ray IDs and span links that correlate work across actors with multi-day lifetimes — a meaningful observability addition for production deployments where a single user request might span many long-lived actors.

  3. 12d ago

    Rivet BYOC: run the control plane in your own AWS or GCP VPC

    ⚡ SPARK

    BYOC (Bring Your Own Cloud) lets enterprise customers run the Rivet control plane inside their own AWS or Google Cloud VPC, fully managed by Rivet with no inbound access from Rivet Cloud. This removes the data residency barrier for regulated industries and directly unlocks enterprise segments that couldn't use a hosted Rivet.

  4. 17d ago

    Rivet MCP: connect Claude Code, Cursor, and Codex to your Actors

    Rivet MCP lets Claude Code, Codex, Cursor, Gemini CLI, and VS Code connect directly to Rivet — listing actors, calling actions, and opening the Actor Inspector without leaving the AI client. Useful developer tooling integration, but primarily serves as distribution: developers using AI coding environments now discover Rivet during agentic coding sessions.

  5. 24d ago

    Durable Streams now supports Rivet Actors

    Durable Streams on Rivet Actors brings real-time pub/sub to the same self-hostable, open-source actor infrastructure — combining stateful computation with event streaming under one abstraction rather than requiring a separate message broker.

  6. 27d ago

    Introducing Dynamic Apps: Deploy AI-Generated Apps for Your Users

    ⚡ SPARK

    Dynamic Apps introduces a V8 isolate runtime for deploying user-generated AI applications inside a developer's backend — scales to zero, no infrastructure management, and designed specifically for the use case where the application itself is generated at runtime by an AI model.