Run the whole live video path

Bring video in, shape it live, route it to the right edge, play it securely, keep every moment, and understand what happened. All through one open control plane.

MistServer media. Federated delivery. Built-in viewer telemetry. Hosted, hybrid, or on your own edges.
Start FreeSee pricing
Start with the workflow you have. Keep the infrastructure options open.

FrameWorks Demo

INITIALIZING

Adaptive playback across WASM, WebCodecs, and every protocol

Auto-negotiates transport and codec per device. Fully embeddable.

One continuous system

Follow one stream through FrameWorks

All capabilities

The sources and destinations change. The path stays legible: bring video in, shape it, send it, keep it, and know what happened.

Move a contribution feed through cueing, processing, regional delivery, archive, and operational telemetry.

  1. 01
    Bring it in

    Contribution

    SRT, RIST, RTMP, HLS and MPEG-TS today; physical and local sources are being productized.

    • AvailableSRT · RIST · MPEG-TS
    • ExpandingSCTE-35 + HLS I/O
    • NextSDI · NDI · devices
  2. 02
    Shape it

    Live processing

    Preserve the source, create renditions, add cue metadata, and place work where capacity exists.

    • AvailableTransmux + passthrough
    • ExpandingABR transcoding
    • NextComposition + SSAI
  3. 03
    Send it

    Delivery

    Route each viewer to an eligible edge and fan the same live source out to external destinations.

    • AvailableHLS · DASH · WebRTC
    • AvailableGeo + health routing
    • AvailableMultistream targets
  4. 04
    Keep it

    Media library

    Turn the live buffer into durable recordings, chapters, clips, thumbnails, and replayable assets.

    • Available24/7 DVR
    • AvailableClips + VOD
    • NextScheduled programming
  5. 05
    Know it

    Operations

    Connect media health, routing decisions, player experience, usage, and automation in one view.

    • AvailableRouting telemetry
    • ExpandingViewer QoE
    • NextIncidents + alerts
One control planeDashboard · GraphQL · CLI · MCPHosted · hybrid · your own edges
01 · Bring it in / Shape it

The media path starts with the signal you already have

Connect contribution equipment, browsers, upstream streams, and files. MistServer keeps the live signal close while FrameWorks places, authorizes, and observes the work around it.

Keep the contribution path open

Push over RTMP, E-RTMP, SRT, or WHIP; pull RTSP, RIST, HLS, MPEG-TS, SRT, and Mist sources.

Preserve broadcast meaning

Carry signal metadata through the media path, including the expanding SCTE-35 and HLS input/output work.

Process where it makes sense

Use the media edge for lightweight work and Livepeer-backed compute for heavier transcoding and AI workloads.

Grow into physical production

NDI, SDI, capture devices, PTZ control, composition, and server-side ad workflows join the same managed model as they land.

Ingest and stream docs
LIVE PIPELINECONTROLLED
Sources
SRT · RISTLIVE
RTMP · WHIPLIVE
HLS · MPEG-TSLIVE
SDI · NDINEXT
Media engine
MistServerin-memory live stream
Live operations
Transmux + passthroughLIVE
SCTE-35 + HLS I/OEXPANDING
ABR processingEXPANDING
Composition + AINEXT
FrameWorks places, authorizes, observes, and automates the workflow.
02 · Send it

One live source, routed to every destination

Fan out to partner platforms and serve viewers from healthy, eligible edges. Routing, failover, playback policy, and delivery telemetry stay part of the same decision.

your encoderMistServerorigin nodeTwitchYouTubeKickFacebookXcustom RTMP / SRTno relay in betweenpushed from hereyour encoderMistServerorigin nodeTwitchYouTubeKickFacebookXcustom RTMP / SRTstart and stop with the stream
One ingest, every destination, pushed straight from the origin. Targets start and stop with the stream.

Native RTMP and SRT push

MistServer pushes from the origin node itself, so there is no relay hop, no added latency, and no proxy bandwidth bill.

Platform presets

Twitch, YouTube, Facebook, Kick, and X are built in. Paste a stream key, or point at any custom RTMP, RTMPS, or SRT endpoint.

Automatic lifecycle

Targets activate when the stream goes live and stop when it ends, with status, error reporting, and a per-target toggle.

Multistreaming docs
Routing is observable

See viewer demand, eligible clusters, and the path chosen for each session, rather than merely a CDN hostname.

03–05 · Play it / Keep it / Know it

Live, replay, and insight stay on one timeline

Playback and recording are not separate products, and analytics is not a detached reporting layer. The same stream identity connects delivery, viewer experience, durable media, and replay behavior.

03 · Play it

Playback is part of operations

Adapt to the browser, enforce access, and return viewer evidence to the same route and edge that served the session.

One adaptive player

HLS, DASH, WebRTC, WebCodecs, WASM, and progressive playback selected per browser and stream.

Access at the boundary

Public, JWT, or webhook authorization for live streams, recordings, clips, and VOD.

Close the delivery loop

Startup, buffering, bitrate, frame drops, geography, and routing decisions share one session timeline.

Playback docs
04 · Keep it

The live stream becomes durable media

Move from the live buffer to seekable replay, durable chapters, clips, and VOD without creating a second workflow.

Record continuously

Keep bounded live seekback while long-running recordings finalize into replayable chapters.

Cut what already exists

Create clips from live buffers, DVR windows, or finalized media through one artifact lifecycle.

Retain with intent

Apply asset policies, freeze durable copies, and understand the storage consequence.

Recording docs
Stream intelligenceLive + replay
05 · Know it

One evidence model, not a separate analytics story

Viewer, route, rendition, and asset context survive from the first frame to the most-replayed moment.

Live viewers12,847+8.2%
Peak concurrent18,402last 24h
Total views284K+3.1%last 24h
Viewer hours39.6Klast 24h
Delivery evidence

What viewers actually experienced

Content evidence

What audiences kept watching

Live bufferDVR chaptersClips + VOD
0:005:0010:00
audience retention most replayed (watch density)
Startup trace

See where the first frame spent its time

608ms to first frame91% cache hit
  • Gateway resolveGraphQL playback resolve48 ms
  • Mist hydratejson_<stream>.js94 ms
  • Player selectProtocol scoring17 ms
  • ConnectTransport + first byte131 ms
  • PrebufferFill to first frame318 ms
06 · Operate it

The controls follow the same system

People, applications, and agents operate the same tenant-scoped resources. The deployment can change without changing the product you build on top.

One control plane

Streams, assets, routing, access, analytics, and billing share the same tenant-scoped GraphQL model.

For people and agents

Use the dashboard, CLI, SDKs, or MCP. Skipper works from the same documentation and operational signals.

Your infrastructure or ours

Use managed regions, enroll your own media edges, or self-host. The application-facing model stays the same.

HostedHybridSelf-hostedEdge nodesControl planeProcessingStorageyour edgeshosted burstControl planeProcessingStorageEdge nodesControl planeProcessingyour S3FrameWorks runs ityou run itexternal, yoursHostedEdge nodesControl planeProcessingStorageHybridyour edgeshosted burstControl planeProcessingStorageSelf-hostedEdge nodesControl planeProcessingyour S3FrameWorks runs ityou run itexternal, yours
Same control plane and dashboards in every row. Move between rows without rebuilding your setup.
import { Player } from '@livepeer-frameworks/player-react'

export const MyStream = ({ playbackId }) => (
  <Player
    contentId={playbackId}
    contentType="live"
    theme="tokyo-night"
    options={{
      autoplay: true,
      muted: true,
      gatewayUrl: "https://bridge.frameworks.network/graphql"
    }}
  />
)
Skipper

I want to go live with a stream called Product Demo

Created stream
verified

"Product Demo" is live-ready. You can go live right now using StreamCrafter in the dashboard, or connect an external encoder like OBS.

I'll use OBS

verified
In OBS go to Settings → Stream → Custom.

Server: rtmp://edge-ingest.frameworks.network/live
Stream Key: sk_live_████████████

Hit Start Streaming and you're live.

I'm live! Can you clip the last 30 seconds?

Clipped 30s
verified

Clipped. "Product Demo Clip 1" is saved to your recordings. Share it directly or download from the stream page.

FrameWorks API, Stream Management
Ask about streams, codecs, or diagnostics...
Agent-native operationsDiscover via SKILL.md, authenticate, fund, inspect, and operate through the same platform APIs.

Discover

Via SKILL.md, MCP, DID, or OAuth metadata.

Authenticate

Wallet signature or x402 creates the account instantly.

Pay

x402 USDC, crypto deposit, or card.

Operate

Streams, diagnostics, and billing all run through MCP.

FAQ

The complete path, plainly answered

Short answers about the media engine, control plane, deployment boundary, and developer surface.

FrameWorks is sovereign live video infrastructure for teams that need hosted, hybrid, or self-hosted streaming without giving up operational control. It combines MistServer delivery, Livepeer-backed processing, GraphQL APIs, QoE analytics, and agent-accessible operations in one platform.

FrameWorks lets you keep the streaming control plane and edge infrastructure in your own footprint while still using hosted load balancing, shared processing, or managed support when you want it. Cloud-only platforms abstract operations away; FrameWorks exposes the machinery so operators can inspect, route, automate, and self-host it.

Yes. FrameWorks can run on bare metal, VMs, or Kubernetes with self-hosted ingest, playback, routing, analytics, and control-plane workflows. Current production deployments still use S3-compatible object storage and managed DNS integrations; native Ceph-backed storage and self-hosted/Anycast DNS are on the roadmap.

MistServer powers low-latency ingest, protocol handling, and delivery at the edge, while Livepeer provides decentralized video processing capacity for workloads such as transcoding. FrameWorks adds tenancy, routing, analytics, billing, automation, and a developer API around those video primitives.

Sovereign streaming infrastructure is useful for broadcasters, community platforms, agencies, events, regulated teams, and builders who need predictable video operations, data control, or deployment freedom. It is especially valuable when a team cannot rely on a single third-party cloud video vendor for every audience, region, or compliance requirement.

AI agents can discover FrameWorks through llms.txt, skill.json, and .well-known MCP metadata, then use wallet authentication, x402 payments, MCP tools, or GraphQL to operate streams and inspect diagnostics. The same platform APIs are available to human operators and autonomous tooling.

Pricing

Transparent Pricing

Compare plans

Start free with self-hosting. Upgrade for advanced processing, hosted services, and enterprise support when you need more.

Backed by Livepeer

Free Tier

Free

Complete self-hosting stack with shared pool access. Open source with permissive licenses: deploy it anywhere.

All self-hosted features
Shared bandwidth pool
Livepeer-backed compute
Open source & permissive licenses
No proprietary video-cloud dependency
Web dashboard with analytics included
Start Free

No credit card required · Deploy in minutes

Agent-Ready

Pay As You Go

€0.00055/delivered min

Pay only when you use it. Wallet-friendly, with account controls for operators.

No subscription, no commitment
Wallet auth with no email or signup form
Top up via card, crypto, or gasless USDC
Same rates as subscription tiers
How wallet pay works

No subscription · No minimum · Agent-native

Paid plans

Hybrid & Hosted

€50+/month

Advanced processing like AI workloads and multi-stream compositing, plus hosted services and enterprise support.

Custom subdomains and hosted load balancers
Reserved processing capacity and bandwidth pools
Team collaboration and advanced analytics
Priority support with 24/7 options
Compare paid tiers

Supporter · Developer · Production · Enterprise

Next steps

Three ways to ship

Deploy the full stack yourself, point your agent at skill.json, or let us run everything. Your call.

Start Free
Browse Docs
Talk to our team