Open source · Pure C · Zero dependencies

Analytics and graphs.
One fast engine.

Rayforce fuses columnar analytics, graph traversal, and recursive queries into one embeddable execution pipeline—built for teams that measure latency in microseconds, not meetings.

Download Rayforce CloudSoon Docs

market_demo.rfl real engine run

; 100,000 deterministic market rows
(select {from: trades
         where: (> qty 100)
         by: symbol
         trades: (count qty)
         notional: (sum (* price qty))})

Loading 100,000 rows

Load100K rows
Optimizerewrite DAG
Filter60K match
Group4 symbols

symboltradesnotional

NVDA15,0001,181,242,500

GOOG15,0001,191,143,000

AAPL15,0001,176,404,200

MSFT15,0001,185,774,600

Recorded with ./rayforce -t 1 Run Rayforce live ↗

Join the community

GitHub X / Twitter Reddit Zulip Email

One execution model

Stop moving data
between engines.

Relational operators and graph traversals belong in the same plan. Rayforce sees the whole workload, rewrites it together, and keeps the hot path close to the metal.

Explore the execution pipeline ↗

Compose

Tables + graphs + rules

Lazy DAG

Optimize

Rewrite the whole plan

Multi-pass

Fuse

Remove intermediates

Bytecode

Execute

Stream cache-sized morsels

Parallel

16Klines of focused C

0external dependencies

IPCclient/server transport built in

MITlicensed and embeddable

Rayfall language

Ask complex questions.
Keep the syntax small.

Use Rayfall interactively, embed the C API, or connect through a growing set of client interfaces. Every surface reaches the same optimizer and execution core.

One engine, three surfaces

; Aggregate high-value flow, then traverse counterparties
(set flow
  (select {from: trades
           where: (> Notional 1000000)
           by: Counterparty
           Volume: (sum Notional)}))

(.graph.var-expand network flow 1 3)

import rayforce as rf

trades = rf.read_csv("trades.csv")
flow = (trades
  .where(rf.col("Notional") > 1_000_000)
  .group_by("Counterparty")
  .agg(rf.col("Notional").sum()))

network.expand(flow, min_depth=1, max_depth=3)

ray_t *trades = ray_read_csv("trades.csv");
ray_graph_t *g = ray_graph_new(trades);

ray_op_t *amount = ray_scan(g, "Notional");
ray_op_t *limit = ray_const_i64(g, 1000000);
ray_op_t *hot = ray_gt(g, amount, limit);

ray_t *result = ray_execute(g, ray_filter(g, trades, hot));

EXAMPLE RESULT5 rows · optimized

CounterpartyDepthVolume

LYNX_02184.20M

ALPHA_17261.08M

NODE_08244.71M

FORT_04327.19M

EDGE_12312.42M

Built as one system

Small footprint.
Serious machinery.

Everything you need to move from raw columns to connected answers—without assembling a second platform around it.

Columnar analytics

Vectorized filters, joins, groups, windows, pivots, and time-series primitives.

↗ 02

Native graph engine

CSR traversal, shortest paths, centrality, communities, WCO joins, and HNSW.

↗ 03

Recursive Datalog

Express reachability and recursive relationships as rules compiled into the same DAG.

↗ 04

Storage that stays open

CSV, splayed and partitioned tables, memory maps, IPC, and block offloading.

↗

In production now

Built for real work,
not synthetic demos.

Rayforce is already part of production systems at teams working with trading, market connectivity, investment analytics, and risk.

Production usage confirmed by the Rayforce project. Customer names link to their public company sites.

PRODUCTION USER · TRADING TECHNOLOGY↗

LYNX

Infrastructure for fast-moving US equity markets.

Lynx Trading Technologies builds modular connectivity, routing, risk, compliance, and analytics systems for broker-dealers.

PRODUCTION USER · INVESTMENT TECHNOLOGY↗

FORTITUDO

Quantitative tools for investment decisions.

Fortitudo Technologies develops investment management software for portfolio analytics, risk, and research workflows.

Rayforce Cloud Coming soon

The engine you can embed.
The platform you won’t have to operate.

We’re bringing Rayforce’s unified analytics and graph pipeline to a managed cloud experience—so teams can move from local prototype to production workload without rebuilding the data path.

Follow development ↗ No launch date announced yet.

RAYFORCE / CLOUDPREVIEW

Deployment

eu-westReady

QUERY SERVICEPreview

ENGINEManaged

Open at the core

Read every line.
Own every workload.

One codebase. One public header. No external runtime. Clone Rayforce, compile it, embed it, and keep control of your data path.

View on GitHub Read the docs ↗

—GitHub stars

—forks

Linux · macOSsupported today