Platform capability inventory
This broader inventory includes the product capability matrix and the infrastructure, trust, and orchestration capabilities behind it. Rows marked (platform) are operated capabilities, not customer-facing features.
Streaming
Section titled “Streaming”| Capability | Availability | Surfaces | Description |
|---|---|---|---|
| Live streams | Available | API, Agents (MCP), Dashboard, Docs | Push streams with stream keys over RTMP/E-RTMP, SRT, and WHIP (WebRTC); pull streams from RTSP, SRT, RIST, HLS, MPEG-TS, and Mist/DTSC sources; and operator-managed local file, playlist, and process sources. |
| ↳ Stream keys (foundation) | Available | API, Agents (MCP), Dashboard, Docs | Per-stream ingest credentials with rotation. |
| ↳ Pull streams | Expanding | API, Agents (MCP), Dashboard, Docs | Stream inputs pulled from upstream sources (RTSP, SRT, RIST, HLS, etc.) instead of pushed, with resolution-stage lifecycle audit. |
| ↳ Multistream push targets | Available | API, Agents (MCP), Dashboard, Docs | Forward live streams to RTMP/SRT destinations with per-target video choice and an optional per-stream live video ABR switch. |
| StreamCrafter (browser studio) | Available | API, Dashboard, Docs | Browser-based broadcasting studio — WHIP/WebRTC ingest, compositing, and multi-source mixing. Ships as npm packages @livepeer-frameworks/streamcrafter-{core,react,svelte,wc} and powers the dashboard’s Go Live page. |
| Device discovery and contribution inputs | Planned | — | Edge-local discovery for cameras, PTZ devices, NDI sources, USB capture, HDMI inputs, and local contribution paths. |
Playback
Section titled “Playback”| Capability | Availability | Surfaces | Description |
|---|---|---|---|
| Playback | Available | API, Agents (MCP), Dashboard, Docs | Resolve a viewer or ingest endpoint for a stream / VOD / clip. Delivery over HLS, LL-HLS (CMAF), DASH, WebRTC (WHEP), and progressive MP4. |
| ↳ Player SDKs | Available | API, Dashboard, Docs | Embeddable player packages — @livepeer-frameworks/player-{core,react,svelte,wc} — with theming, adaptive playback, and built-in boot/QoE telemetry. |
| ↳ Viewer protocol selection | Available | API, Docs | resolveViewerEndpoint accepts a required playback protocol and never substitutes another format. Players check serverInfo for this slug before sending the argument. |
| ↳ Playback access control | Available | API, Agents (MCP), Dashboard, Docs | Public / JWT / webhook playback authorization policy on streams, clips, and VOD assets, with allowed embedding origins, a per-policy webhook context object, a dry-run tester, and full agent management surface. |
| ↳ Signing keys (foundation) | Available | API, Agents (MCP), Dashboard, Docs | ES256 keypairs for signing playback JWTs. Private material returned once on creation. |
| ↳ DRM | Planned | — | FairPlay, Widevine, and PlayReady protection for premium playback workflows. |
Media Library
Section titled “Media Library”| Capability | Availability | Surfaces | Description |
|---|---|---|---|
| DVR recordings | Available | API, Agents (MCP), Dashboard, Docs | 24/7-capable DVR recording with bounded live seekback, continuous archive storage, chapter replay, and VOD-style playback. |
| ↳ DVR chapters | Expanding | API, Agents (MCP), Dashboard, Docs | Chapter artifacts that segment a long-running DVR session into replayable windows, so a 24/7 recording can be replayed by window, civil-time range, or explicit range without restarting the live stream. |
| ↳ Scheduled recordings | Planned | — | Calendar-driven DVR captures that start and stop on schedule, so streams record themselves on a fixed cadence. |
| VOD assets | Available | API, Agents (MCP), Dashboard, Docs | Multipart VOD upload, import from a public http(s) URL, and lifecycle. |
| ↳ Scheduled VOD streams | Planned | — | Linear live channels assembled from existing VOD assets, with schedules that make archived media play as a rolling live stream. |
| ↳ Upload validation and presign hardening | Expanding | — | Enforce declared size, content type, and container validity on presigned VOD uploads. |
| Clips | Available | API, Agents (MCP), Dashboard, Docs | Time-range clips from live streams, DVR windows, and finalized chapters — stored as their own playable assets via createClip, or cut on the fly as a single MP4/MKV/WebM/TS download by adding time-range parameters to a playback URL. |
| Media retention | Available | API, Agents (MCP), Dashboard, Docs | Tenant-controlled storage policy for 24/7 DVR archives, clips, and VOD assets, with per-asset overrides and entitlement bounds. |
| Thumbnails and sprite previews | Expanding | API, Agents (MCP), Dashboard, Docs | Auto-generated poster frames and sprite-tile preview strips for live streams, clips, VOD, and DVR playback. |
Processing & AI
Section titled “Processing & AI”| Capability | Availability | Surfaces | Description |
|---|---|---|---|
| Processing (foundation) | Expanding | API, Dashboard, Docs | Server-side media processing — live ABR transcoding, on-edge AI, and composition — placed on the right nodes across a cluster. |
| ↳ Gateway-backed transcoding | Expanding | API, Dashboard, Docs | Live ABR renditions through Livepeer Gateway or local processing, with processing usage telemetry. |
| ↳ Live AI processing | Planned | — | ONNX models running directly on the in-memory stream buffer of media edge nodes — detection, transcription, moderation, audio tagging, and embeddings for semantic search. |
| ↳ Composition and SSAI | Planned | — | MistServer-native blitter for logos, watermarks, and lower-third overlays; Composer for server-side ad insertion (SSAI), raw-pixel composition, picture-in-picture, and automated clipping. |
| Workload cost model (GPU-aware scheduling) (foundation) | Planned | — | Predictive per-job resource cost — codec, resolution, framerate, model size, and realtime factor mapped to expected CPU, GPU/VRAM, and bandwidth — continuously corrected by measured costs of completed jobs, feeding processing placement and admission. |
Analytics & Observability
Section titled “Analytics & Observability”| Capability | Availability | Surfaces | Description |
|---|---|---|---|
| Analytics | Expanding | API, Agents (MCP), Dashboard, Docs | Stream health, viewers, geographic, routing, quality tiers, federation, node metrics, and view-level player QoE (rebuffering, frame drops, startup, VOD retention) — with cross-player coverage still expanding. |
| Incidents and alerts | Available | API, Agents (MCP), Dashboard, Docs | Incidents built from platform alert rules, scoped to platform operators or to the tenant that owns the alerting cluster, with acknowledge, assign, resolve, notes, operator email/Slack/Discord notifications, realtime tenant updates, and Skipper investigation of tenant incidents. |
Infrastructure & BYOC
Section titled “Infrastructure & BYOC”| Capability | Availability | Surfaces | Description |
|---|---|---|---|
| Geo-balanced viewer routing (foundation) | Available | API, Agents (MCP), Dashboard, Docs | Foghorn resolves each viewer to the optimal edge by geography, load, and node health across federated clusters. |
| Network egress economics and peering (foundation) (platform) | Planned | — | Operator-side supply-cost attribution, destination-network traffic analysis, percentile demand forecasting, and evidence-gated transit or peering evaluation for media egress. |
| Edge clusters (foundation) | Available | API, Agents (MCP), CLI, Dashboard, Docs | Tenant-owned MistServer/Helmsman edge capacity with bootstrap and node enrollment; virtual private clusters are logical tenant-private slices on provider-operated Foghorn. |
| ↳ Enrollment / bootstrap tokens (foundation) | Available | API, Agents (MCP), Dashboard, Docs | One-shot tokens used to enroll new edge nodes into a cluster. |
| ↳ Nodes (edge / ingest / playback) (foundation) | Available | API, Agents (MCP), CLI, Dashboard, Docs | Per-node health, info, and discovery. |
| ↳ Node mode (drain / maintenance) (foundation) | Available | API, Agents (MCP), Dashboard, Docs | Operator controls to drain a node, put it in maintenance, or restore it to active service. |
| ↳ Federated network status (foundation) | Available | API, Dashboard, Docs | Cross-cluster routing visibility and orchestrator vantage points. |
| ↳ Cluster ingest protocol allowlist (foundation) | Planned | — | Per-cluster allowlist of ingest protocols (WHIP, RTMP, SRT) enforced at placement and publisher admission. |
| ↳ Regional edge redundancy (foundation) | Planned | — | More than one edge node per region per cluster, so a single node failure does not take a region’s media offline. |
| ↳ Edge auto-update (foundation) | Available | CLI, Docs | Automatic edge component updates — Quartermaster owns the release catalog and per-cluster targets; Foghorn’s release reconciler applies them to Helmsman, MistServer, and Caddy over the Helmsman control stream — native and single-image container nodes alike. |
| ↳ Edge ISO and cloud images (foundation) | Planned | — | Bootable FrameWorks Edge ISO and prebuilt cloud/marketplace images for on-demand media-cluster scaling. |
| Cluster marketplace | Expanding | API, Agents (MCP), Dashboard, Docs | Browse listed clusters and manage subscriptions. A vetted public marketplace program for third-party capacity is on the roadmap. |
| ↳ Cluster invites (foundation) | Available | API, Agents (MCP), Dashboard, Docs | Direct invites granting another tenant access to a cluster. |
| ↳ Cluster subscription requests (foundation) | Available | API, Agents (MCP), Dashboard, Docs | Tenant-requested subscription to a cluster, requiring operator approval. |
| BYOC ownership levels | Expanding | CLI, Docs | The three supported ways to own a FrameWorks deployment — hosted, self-hosted edges attached to the FrameWorks platform, or the entire stack self-hosted — and the one unsupported combination between them. |
| ↳ Federating a self-hosted platform with FrameWorks | Not available | — | Run your own Foghorn and control plane and federate them with the FrameWorks platform as one network, sharing routing, capacity, and tenants across both operators. |
| ↳ Hosted private clusters (BYOC) | Planned | — | FrameWorks-provisioned dedicated, isolated clusters for a single tenant, with reserved capacity and fixed-rate pricing. |
| ↳ Full self-host | Available | CLI, Docs | Run the entire public-domain stack — control plane, Foghorn, media edges, Livepeer, Chandler, databases, analytics, and interfaces — as an independent platform, provisioned, upgraded, and backed up with the frameworks CLI from a GitOps repository. |
| Adjacent infrastructure services | Expanding | API, Docs | Infrastructure products adjacent to video delivery — managed DNS, generic proxying/CDN, and standalone storage. |
| ↳ Managed DNS and cluster domains | Expanding | API, Docs | Platform-managed DNS, TLS certificates, media-cluster domains, and tenant streaming domain discovery. |
| ↳ Generic proxying and CDN services | Planned | — | Customer-configured proxy/CDN front doors for non-video services and adjacent workloads. |
| ↳ Standalone object and block storage | Planned | — | Customer-facing storage buckets, volumes, policies, and lifecycle controls outside the video-asset workflow. |
| NAT-aware routing and traversal | Planned | — | NAT-type classification as a first-class routing and placement signal, platform-coordinated hole punching, and platform-managed measurement of inter-server network performance (latency, throughput) feeding routing and capacity decisions — so self-hosted nodes behind NAT participate fully in the federated network. |
| Federation state synchronization | Planned | — | Decentral distribution of configuration and state changes across a dynamic federation — clusters and servers join, leave, or drop offline at will — with automatic conflict resolution between concurrent changes and reconvergence after network partitions, without a central coordination point. |
| Stream replication orchestration | Planned | — | Pre-emptive stream distribution — spread a stream to a cell or region ahead of demand, explicit hop topology, and per-stream replication policy (max replicas, allowed regions). |
| Ingest and viewer placement policy | Expanding | API, Agents (MCP), Dashboard, Docs | Ingest and viewer placement policy with cluster-owner permissions, authorized selectors, read-only previews, reviewed changes, and save recovery, enforced on live routing and at final admission. |
| Placement policy and storage tiering | Planned | — | Two-sided placement policy — cluster-owner and content-owner constraints (data residency, minimum durable copies) compiled into the signed media authority evaluated at serve/ingest/process/store/replicate/peer decision points — over a cold-to-hot storage tier ladder with a backend-keyed placement ledger. |
| Drain-integrated OS updates (foundation) (platform) | Planned | — | Operator-driven host OS updates that drain each role safely before restart — media sessions run out or hand over, database replicas gate on recovery state, brokers transfer leadership — while cluster capacity floors hold. |
| Dynamic mesh routing (foundation) (platform) | Planned | — | Dynamic route propagation over the encrypted WireGuard mesh so the topology adapts to link failure and growth without quadratic static peer configuration. |
| Service identity and trust lifecycle (foundation) (platform) | Planned | — | Live rotation of trust material without service interruption — dual-validity intermediate CA rotation, asymmetric token authority (the signing key never leaves the issuing service), and non-forgeable cluster-bound service identities. |
| Foghorn high availability (foundation) (platform) | Available | — | Regional balancer HA — Redis-externalized state, leader election, and changelog sync per cell so any Foghorn instance can serve routing decisions and node-mode changes propagate immediately. |
| Trigger durability WAL (foundation) (platform) | Available | — | Edge-local write-ahead log for billing-critical media triggers — acknowledged only after durable append, deterministic event identity for idempotent replay, and an operator replay endpoint on the edge sidecar. |
| Internal CA and certificate sync (foundation) (platform) | Available | — | Two-tier internal PKI — 72-hour service leaves issued by Navigator’s internal CA via node- and cluster-bound cert_sync tokens, distributed by Privateer; public ACME (DNS-01) covers external listeners. |
| WireGuard mesh and internal DNS (foundation) (platform) | Available | CLI, Docs | Self-managed encrypted service mesh with no external dependency — three-layer peer configuration (gitops seed, last-known, managed), dependency-derived topology, and a loopback .internal resolver per node. |
| Cluster rollout orchestrator (foundation) (platform) | Available | CLI, Docs | Manifest-driven convergence — dependency-ordered phases, canary and region-staggered waves with health gates, content-addressed release pinning, and automatic rollback on failed health validation. |
| Cluster backup and restore (platform) | Available | CLI, Docs | CLI backups of every authoritative store — per-database PostgreSQL/YugabyteDB dumps and the ClickHouse billing facts — to a local directory or S3, with a manifest of file hashes, row counts and migration-ledger digests. Restores go through a verified shadow and a reversible swap; contract and irreversible data migrations require a backup under an hour old that still matches the live ledgers. |
| Livepeer transaction signer (foundation) (platform) | Available | — | Keystore-backed remote ETH transaction signer so the Livepeer gateway never holds key custody on media nodes. |
Commerce & Billing
Section titled “Commerce & Billing”| Capability | Availability | Surfaces | Description |
|---|---|---|---|
| Billing & prepaid balance | Available | API, Agents (MCP), Dashboard, Docs | Card / Mollie / crypto top-ups, billing details, invoices, prepaid balance. |
| ↳ Usage caps, alerts, retention cost preview | Expanding | — | Customer-set caps and alerts on storage / bandwidth / recording / transcode meters; cost-preview hypotheticals for retention policy changes. |
| Wallet & on-chain identity | Expanding | API, Agents (MCP), Dashboard, Docs | Wallet sign-in, linked-wallet account identity, and x402 micropayments. |
| Operator settlement and attribution | Planned | — | Attribute served traffic to the operators whose clusters actually carried it — manipulation-resistant across independent operators — and settle the full loop, from credit accrual through operator-visible reporting to payout execution. |
| Stream-level balances | Planned | — | Optional per-stream balances that fund usage before the tenant balance — pay-per-view, tips, sponsor-funded streams — plus promo codes, referral rewards, and volume discounts. |
| Operator credit ledger (foundation) (platform) | Expanding | — | Internal accrual ledger crediting cluster operators for served usage — platform-fee basis points, clawback and reversal support — populated from per-cluster usage attribution. |
Developer Platform & Agents
Section titled “Developer Platform & Agents”| Capability | Availability | Surfaces | Description |
|---|---|---|---|
| Developer API | Expanding | API, Agents (MCP), Dashboard, Docs | The GraphQL API and the tooling around it — playground, API tokens, webhooks, and server-side SDKs. |
| ↳ Developer API tokens (foundation) | Available | API, Dashboard, Docs | Long-lived API tokens for SDK / CI use. |
| ↳ GraphQL playground | Available | Dashboard, Docs | Interactive GraphQL explorer wired to the live schema. |
| ↳ Public tenant events | Available | API, Dashboard, Docs | One public event contract for streams, clips, recordings, uploads, multistream targets, API tokens, billing, account, and custom domains, written by each owning service in the same transaction as the change and delivered live through the tenantEvents subscription, filterable by type and stream. |
| ↳ Outbound event webhooks | Available | API, Agents (MCP), Dashboard, Docs | Up to 10 tenant HTTPS endpoints receive the public tenant events (the tenantEvents payloads) as Standard Webhooks-signed requests, retried for about three days with auto-disable after sustained failure, with a 30-day delivery log, single and range replay, test deliveries, and secret rotation with a 24-hour overlap. Delivered by the Bosun service. |
| ↳ Backend SDKs | Expanding | Dashboard, Docs | SDKs for TypeScript (Node 22+ and browsers), Go, and Python, generated from the public GraphQL schema and released under one version. Each has a generated typed method, with a default selection, for every non-deprecated public root query, mutation, and subscription (a namespace root such as analytics through the methods of its argument fields) and every non-deprecated argument-taking field in the public schema. Argument-free fields nested deeper than the default selection are typed only through the TypeScript select client; Go and Python reach them with custom documents (Go MakeRequest, Python execute). Around them are retries, pagination, typed errors, a server version check, a multipart VOD upload helper, playback JWT signing, and Standard Webhooks verification with typed event parsing. |
| Agents (MCP server) | Available | API, Agents (MCP), Dashboard, Docs | Model Context Protocol server exposing FrameWorks operations to agents. |
| ↳ Skipper (in-app consultant agent) | Expanding | API, Agents (MCP), Dashboard, Docs | Conversational platform agent for support and investigations. |
Engagement & Interactivity
Section titled “Engagement & Interactivity”| Capability | Availability | Surfaces | Description |
|---|---|---|---|
| Conferencing and remote contribution | Planned | — | Multi-party WebRTC rooms and remote-guest contribution — commentators and callers mixed into a managed stream. |
| Rooms, chat, and interactivity (Parlor) | Planned | — | Viewer chat and engagement primitives — channel points, hype trains, leaderboards, flair, and moderation views — synced to the stream. |
| Live commerce (shopping and auctions) | Planned | — | Shopping and server-authoritative auctions attached to live streams — product showcases, carts and orders, and auction mechanics that stay fair for viewers at different latencies behind the live edge. |
Account & Apps
Section titled “Account & Apps”| Capability | Availability | Surfaces | Description |
|---|---|---|---|
| Tenant settings | Expanding | API, Agents (MCP), Dashboard, Docs | Tenant identity, billing setup, retention defaults, routing preferences, and wallet-provisioned account lifecycle. |
| ↳ Teams & RBAC | Planned | — | Multi-user tenants with role-based access control. |
| Companion apps | Expanding | — | Desktop and mobile companions to the dashboard and CLI. |
| ↳ Desktop companion (macOS tray) | Expanding | — | Menu-bar app bundled with the CLI in the macOS installer for local edge and stream control. |
| ↳ Mobile apps and integrated help center | Planned | — | Native mobile operations and an in-product help center for support, incidents, and account workflows. |
| Platform-operator god view (foundation) (platform) | Available | — | Cross-tenant operational view for platform operators — audit-logged and read-oriented — used for support, incident response, and platform-wide analytics. |
Product-facing sequencing lives on the Roadmap; this inventory is regenerated from docs/platform-features.yaml by make generate-feature-registry.