Summary: Apibara is ingest plumbing with an unusually legible wire protocol. DNA pushes filtered chain data, invalidations, finality signals, and liveness over gRPC, so the note is worth keeping for the stream boundary, not for another indexer SDK story.
What it does:
Defines and implements the DNA protocol for streaming onchain data directly from nodes or node-adjacent servers into applications over gRPC
Ships a developer runtime and CLI for building TypeScript-based indexers that backfill history, switch to live mode, and handle reorgs automatically
Supports chain-specific streams and schemas, with current public emphasis on EVM and Starknet data plus Starknet appchain ingestion
Lets developers target either Apibara-hosted streams or self-hosted DNA endpoints through configurable streamUrl inputs
Exposes plugins, hooks, and storage integrations as optional layers rather than baking one persistence model into the protocol itself
Positions itself as open-source infrastructure that can run locally, on appchains, or in managed production deployments
Key claims:
The homepage’s strongest framing is Direct Node Access: the node reads data from its database and sends it directly to the application through gRPC. That is the clearest reason to file Apibara as an ingest protocol plus runtime, not just a hosted indexing service.
DNA is the real mechanism note. Streams use cursors that carry both order and branch identity, so clients can resume cleanly and detect when a previous head was invalidated by a reorg.
DNA’s wire messages matter analytically: data, invalidate, finalize, heartbeat, and system_message make reorgs, liveness, and server policy visible instead of burying them inside SDK retries.
The indexer docs keep protocol and application logic more separate than most indexing vendors do. Developers choose filters, finality mode, runtime config, plugins, and optional reorg hooks, while storage stays pluggable.
The practical choke point is whoever runs the DNA endpoint. The protocol may stay the same, but the operator still chooses retention, schema exposure, backfill behavior, and runtime defaults.
That makes Apibara a useful contrast to the stronger ingest anchors in this branch: it keeps the ingest boundary visible instead of hiding it inside a broader data product.
Whitepaper: No classic whitepaper surfaced in this pass. The strongest primary materials were the official site, hosted docs, the documentation corpus, the DNA protocol docs, and the canonical DNA implementation repository; see ../whitepapers/apibara-primary-sources-2026-05-14.md.