solana-indexer

fills the indexing-geyser gap

Every Solana product
runs on an indexer.

Polling RPC is a tax. Rate limits are a ceiling. The AI Kit had skills for using indexed data — nothing for building the thing that produces it. This fills that gap.

View bounty PR → Full repository

tsc

TS example

cargo

Rust example

9 / 9

refs resolve

6

CI jobs

35

files, MIT

01 — The problem

Every dApp needs a custom indexer.

DeFi dashboards, NFT marketplaces, gaming leaderboards, social graphs, analytics — every serious Solana dApp needs a custom indexer. The AI Kit had Helius, QuickNode, and solana-agent-kit skills for consuming indexed data. Nothing for building one. Every founder had to rediscover the same patterns.

02 — What works

Everything in the box.

9 reference docs

indexer-architecture, geyser-plugins, postgres-schemas, backfill-strategies, real-time-streaming, cost-optimization, testing-indexers, production-ops, resources.

3 runnable examples

minimal-indexer-ts, geyser-plugin/skeleton, subgraph-template. All type-checked.

2 agents + 2 commands

indexer-architect, indexer-qa, /build-indexer, /backfill. 1 rule auto-loads in /indexer folders.

Progressive loading

SKILL.md is 3.9 KB. References load only when the agent needs them. Idle context cost: <4 KB.

03 — Install

One line.

install.sh 
$ curl -fsSL https://raw.githubusercontent.com/srivtx/solana-indexer-skill/main/install.sh | bash

Installs to ~/.claude/skills/solana-indexer/ (and ~/.codex/ if installed). Restart Claude Code or Codex.

04 — In action

Watch it work.

Demo: asking Claude Code how to index Raydium CLMM swaps — the skill loads, routes to the right reference, and the Rust Geyser plugin compiles and runs against Postgres

Asking Claude Code how to index Raydium CLMM swaps. The skill loads, routes to indexer-architecture.md, recommends Geyser gRPC, verifies the Postgres schema against Raydium’s PoolState source, and the Rust plugin compiles and streams 50K swaps/min on mainnet.