Community-built infrastructure mapped to the Catalyst roadmap — an offer to collaborate
Following last week’s blog post on the Foundation’s new mandate and this week’s Indexer Office Hours walkthrough of Project Catalyst, we wanted to put something concrete on the table.
Over the past months, Night’s Watch has been independently building infrastructure for the network — no grants, MIT/Apache licensed, all public. When we mapped our repos against the Catalyst roadmap items presented on the call, the overlap was substantial. We’re posting this so the Foundation’s incoming technical team doesn’t spend its first quarter rebuilding things that already exist, and so the community can kick the tires on what’s there.
(For the less serious version of this post, see our blog. This one is the sit-down-with-coffee version.)
The mapping
1. Studio continuity / network-powered Studio → Lodestar Dock
Lodestar (repo) already ships a Studio-parity developer surface: full on-chain subgraph lifecycle (metadata updates, ownership transfer, deprecation), deploy keys, an embedded GraphiQL playground with the real gateway query URL, per-subgraph Discord/Slack health alerts, and a non-custodial metered query gateway with self-minted API keys and a free tier.
What it doesn’t do: the DIPS payment plumbing. Making indexers earn for syncing Studio subgraphs is protocol work that belongs with the Foundation’s transition effort — but the frontend/UX layer doesn’t need to be rebuilt from zero, and we’re happy to align the Dock with whatever the official Studio migration needs.
2. New gateway operators → gib (Gateway-in-a-Box)
gib is a self-hostable, TAP v2 / Horizon-native subgraph gateway in a Docker Compose stack. Measured footprint is ~570 MB for the full stack with the complete Arbitrum One topology resident — it runs on a 2 GB VPS. It ships with a one-command smoke test that proves the entire payment path (topology sync, receipt signing, RAV aggregation, EIP-712 domain, negative tests) with zero third-party cooperation, so a prospective operator arrives at an indexer’s door with verifiable output rather than promises.
We’ll be direct about status, because the README is: no payment has flowed through it yet, and the real onboarding cost is the indexer whitelist (tap.sender_aggregator_endpoints) — a social coordination problem, not a technical one. If the Foundation is courting gateway operators, gib removes the technical setup barrier and the whitelist coordination is exactly where Foundation support would be highest-leverage. We’d welcome a shared onboarding process (or even a lightweight “gateway operator registry”) built around it.
3. AI / agent access → compass (Subgraph-MCP data service)
The “Memory for AI” service described on the call is the Foundation’s own work, and we’re genuinely looking forward to it. Adjacent to it, compass is a decentralised Subgraph-MCP gateway as a Horizon data service: every subgraph an indexer serves becomes a discoverable, typed, pay-per-call MCP tool — settled in GRT via TAP v2/GraphTally, or in USDC on Base via x402 for agent-discovery flows. Any MCP-speaking agent can query subgraphs for sub-cent amounts with no API key and no central operator.
Weeks 1–6 of the build are complete (contract tested and deployed to Arbitrum Sepolia, TAP validation, x402 rail, schema-introspected per-entity tools); the CLI and mainnet launch are next, and a GRC forum post will follow. If the Foundation’s AI push needs an agent-facing access and payment rail, this is intended to be exactly that, and we’d rather design it with the Memory for AI team than in parallel.
4. Substreams data service → SDSCE
SDSCE (Substreams Data Service Community Edition) is a maintained fork of graphprotocol/substreams-data-service with: a live SubstreamsDataService contract on Arbitrum One (UUPS-upgradeable, fixed 1% burn on collected fees), an automated RAV settlement daemon, deployment and onboarding runbooks, an internal security audit plus an external-audit brief, and a rehearsed end-to-end provision → register → collect path against Sepolia Horizon contracts.
We named it “Community Edition” deliberately: it is experimental, externally unaudited, currently EOA-owned, and explicitly leaves room for an official deployment. With StreamingFast supporting the Foundation’s resumed effort, our preferred outcome is not two parallel contracts — it’s handing over the deployment learnings, runbooks, and fork so the official service ships faster. Consider this the formal offer.
5. RPC service → Dispatch (GRC-005)
The call mentioned that an RPC service plan is finally coming together. Dispatch exists under GRC-005 as a JSON-RPC Horizon data service, and it’s mature enough that Lodestar’s indexer scoring already credits multi-service coverage including it. We’d suggest the plan start from evaluating Dispatch rather than a blank page — and we’re open to whatever governance process (GRC revision, audit, handover, co-maintenance) makes it official-track.
6. Multi-product experience
Lodestar already operates as a multi-product surface: indexer directory and scoring, delegator and curator portfolios, one-click delegation, POI consensus, GraphTally/TAP payment tracking, indexing health, the Dock, and an AI/MCP directory. If the Foundation’s multi-product Studio effort wants a reference for what the community actually uses (and where the sharp edges are), our GAP_ANALYSIS and issue tracker are public.
What we’re asking for
Not grants — we’ve gotten this far without them. Concretely:
- A conversation with the incoming product/protocol team (Pedro, Orjan, and colleagues) to walk through the above and identify where adoption or alignment beats rebuilding.
- Foundation support on the gateway whitelist coordination problem — the one piece of gateway operator onboarding no independent team can solve alone.
- A decision path for the community data services (SDSCE, Dispatch, compass): what would it take — audits, governance, ownership transfer — for community-built Horizon data services to become official-track rather than parallel-track? A clear GRC → official pipeline would benefit every builder here, not just us.
- Feedback and issues from indexers, delegators, and developers on any of the above. Everything is public and MIT/Apache licensed.
The new mandate’s core bet is that a smaller, focused team can ship faster. We agree — and we’d point out that some of the shipping has already happened. Let’s not waste it.
— Night’s Watch GitHub: Night's Watch · GitHub · Lodestar: https://lodestar-dashboard.com