Skip to main content
Markdown
DOCUMENTATION

Exaloop

Native performance. Familiar Python syntax.

Build data pipelines, database functions, and streaming applications. Profile real workloads and ship native executables.

Get Exaloop Write your first application

Your systems. One native stack.​

Platform overview

Keep the systems you trust. Bring native execution to the code that connects them.

DataArrow, Parquet, Avro, ProtobufDatabasesDuckDB, PostgreSQL, MySQL, RedisStreamingApache KafkaAI agentsGenerate + optimizeExaloop AI NATIVE EXECUTIONCompiler + native runtimeNative NumPyParallel executionIntegrated SDKsNative applicationsPipelines, services, workersDatabase functionsNative compute, in place
Performance feedback01Measure02Optimize03Rebuild
One toolchain from application code to native artifacts. AI agents generate Codon code and use measured feedback to guide optimization.

INTEGRATIONS

Apache ArrowParquetDuckDBPostgreSQLMySQLRedis

Start with a real workload​

Explore the cookbook
CSV + Parquet

Reconcile invoices and settlements

Turn two exports into a checked, auditable reconciliation report.

DuckDB + Arrow

Screen a loan portfolio in SQL

Run native risk calculations inside DuckDB and publish the results to Parquet.

Kafka + Avro

Process orders in a Kafka transaction

Validate events and commit consumed offsets with the output transaction.

Get started​

01

Install Exaloop

Set up the compiler, libraries, and command-line tools for your platform.

Installation guide
02

Build an application

Create a project that writes and reads a Parquet file with native code.

First application
03

Prepare for production

Package native dependencies and check the destination environment.

Deployment guide

Understand where the time goes​

Follow a workload from function calls to memory use. Share the profiling report, identify expensive paths, and use representative runs to evaluate optimized builds.

Native profiling of a synthetic portfolio-risk workload with 250,000 loans and 96 stress scenarios, showing execution totals, function timings, and memory useNative profiling of a synthetic portfolio-risk workload with 250,000 loans and 96 stress scenarios, showing execution totals, function timings, and memory useExplore profiling

Integration guides​

Files & messages

Move between native values and the formats your systems already use.

Queries & events

Build a client application or bring a calculation into the database.

Performance

Measure execution time and memory, then evaluate profile-guided builds.

Operations and support​