live network: 709 verified empirical runs • 19 calibrated task cohorts
kangarouter kangarouter edge hub v1
provider-neutral mathematical control plane

empirical bayesian routing for ai coding agents

replaces synthetic benchmarks with real-world test suite verification. helps claude, codex, and cursor choose the optimal model for every refactor, debug, and architecture task.

practical workflow

how you actually use kangarouter with claude & codex

if you build daily with claude code, claude desktop, chatgpt codex, or cursor, here is the exact 3-step loop:

STEP 1: 1-CLICK PAIRING

add to claude config

add kangarouter as an mcp server in claude_desktop_config.json:

"mcpServers": {
  "kangarouter": {
    "command": "npx",
    "args": ["-y", "kangarouter", "mcp"]
  }
}
STEP 2: CHAT NATURALLY

ask claude to route

type to claude as you normally would:

"check kangarouter for the best model to rewrite this auth middleware across 5 files"

claude calls kangarouter_get_prior in the background.

STEP 3: EMPIRICAL ASSURANCE

get real win-rates

claude checks real test-pass data across 700+ verified runs:

"for HIGH risk multi-file writes, claude-sonnet-5 has a 94.2% test-pass confidence. estimated duration: 45s."

live network intelligence state

aggregated empirical performance from independent contributing clusters

live d1 • london edge
709
verified runs
empirical test passes
19
task cohorts
shape-specific priors
4
provider clusters
anthropic • openai • google
0.00%
data egress risk
zero code / zero prompts

empirical prior resolver

inspect real candidate distributions for any task shape

differential privacy masked get /v1/route-priors
resolved candidate ranking: bounded_planning|LOW|S|BOUNDED|READ_ONLY
resolving edge priors...

connect your agent fleet in 60 seconds

works out-of-the-box with codex, claude code, cursor, aider, and custom python agents

node.js / npm v0.0.1

run directly inside any workspace:

npx kangarouter connect
python / pypi v0.0.1

install standard python package:

pip install kangarouter