Dedalus
LLMOps workbench for AI architects

Architecting the future of AI. Built for pros.

Dedalus is the infrastructure engineering platform for serious AI teams. Prune monorepo context with AST precision, compose multi-agent workflows, and govern your LLMOps from a single workbench.

Start building freeRead the docs

$ npm i -g @dedalus/cli  ·  no credit card required

dedalus prune ./monorepo

→ scanning 4,182 files across 38 packages

→ building abstract syntax trees…

→ pruning noise · keeping skeleton context

Raw tokens1,284,500
SlimContext86,200

✓ 93.3% reduction · ready for Claude Code

Trusted by AI engineering teams shipping in production

NorthwindHelix AIForge LabsCortexVantailQuanta

93%

median token reduction

12M+

agent runs orchestrated

<40ms

p95 routing overhead

99.99%

control-plane uptime

The workbench

Escape the context labyrinth

Three precision instruments for AI engineering — from raw repository to governed production pipeline.

01Context Pruner

AST-based context pruning

Stop blowing your context window. Dedalus parses your entire monorepo into abstract syntax trees and extracts only the highest-relevance skeleton context — slashing token cost while sharpening AI output.

  • Compress and slice massive monorepos into SlimContext
  • Relevance ranking across files, symbols and dependencies
  • Drop-in for Claude Code, Copilot and custom agents
context.treeAST · pruned
keepmonorepo/
keeppackages/core/auth.ts
prunepackages/core/__tests__/
keeppackages/ui/button.tsx
prunepackages/ui/storybook/*
prunenode_modules/**
keepservices/api/router.ts
prunedocs/legacy/*.md
Skeleton context extracted−93.3%
02Labyrinth Workflow Composer

Compose multi-agent pipelines

Design complex agent systems like building a labyrinth. A visual node canvas for orchestrating pipelines with dynamic routing, conditional branches and precise RAG memory injection.

  • Visual node editor for multi-agent orchestration
  • Dynamic routing and conditional branching
  • Short- and long-term memory (RAG) injection

Input Router

intent · classify

branch

Retriever Agent

RAG · vector

Coder Agent

tools · exec

Critic Agent

eval · score

Memory Store

long-term

Composer Output

merge · stream

03Pro LLMOps Dashboard

Govern your model performance

Real-time observability across every model API you call. Track latency, token throughput, spend and generation quality — then optimize your multi-model routing with one click.

  • Live latency, throughput and cost monitoring
  • Automated generation-quality evaluation
  • One-click multi-model routing optimization
llmops · livestreaming

Throughput

18.4ktok/s

p95 latency

39ms

Spend / 24h

$412

gpt-class · fast
$0.0021
reasoning · pro
$0.0185
local · 8B
$0.0004

Master the LLM infrastructure

Join the engineering teams building the next generation of AI systems on Dedalus. Start free — scale when you ship.

Start building freeTalk to engineering