Weave Router Client Routing Policy Setup (Onboarding)
A sequence with 7 steps: Run npx @workweave/router against the client's Claude Code, Codex, or Cursor install.
By InnovaAI ResearchPublished
What are the steps?
Weave Router Client Routing Policy Setup (Onboarding)
- 01
Run npx @workweave/router against the client's Claude Code, Codex, or Cursor install
The single command detects which coding-agent harness is present and writes the router config. Confirm the complexity classifier responds in single-digit milliseconds before you touch routing rules, because a slow classifier adds latency to every agent turn the client's developers take.
- 02
Load model credentials for the cheap tier and the frontier tier separately
Routine turns should land on DeepSeek, Llama, or Gemini; Claude and GPT stay reserved for escalation. Keep the two credential sets distinct so a quota problem on one tier does not silently disable the other mid-session.
- 03
Set the routing policy that governs when a turn escalates
Define what counts as stalled or looping for this client's workload. The escalation classifier returns looping tasks to frontier models automatically, so the policy you write here decides how much of the client's monthly spend reaches Claude or GPT.
- 04
Enable the cache-aware cost check and subscription quota draining
Turn on quota draining so flat-rate subscription capacity is consumed before the router pays per-token. Verify both switches in a staging session, not on live client traffic, since a misconfigured drain order shows up as unexpected per-token charges on the first invoice.
- 05
Benchmark router quality, speed, and cost against a frontier-only baseline
Run the same representative task set twice, once routed and once pinned to a frontier model, and record token spend for both. This baseline is the number you show the client at the 30-day retainer review, so capture it before routing goes live.
- 06
Build the usage dashboard report showing model spend and routing decisions
The client needs to see which turns escalated and why. A report template that maps spend to routing decisions turns an infrastructure line item into something an agency can defend in a renewal conversation.
- 07
Document the handoff guide and train the client's engineering team
Cover how to read escalation logs, how to adjust thresholds when a sprint shifts toward harder work, and who to contact when routing behaves unexpectedly. Agencies that skip this step absorb every threshold question as unbilled support.
More on Weave Router
- StrategyWhy Weave Router Is a Cost-Control Layer, Not an Agency Retainer
- ConceptWeave Router Escalation Budget
- Evaluation RuleWeave Router Rule: Adopt Only When Your Client's Coding-Agent Token Spend Exceeds the $1,800 Setup Fee
- Decision FrameworkWeave Router: Buy vs Skip (Agency AI Coding Spend)
- Failure PatternThe Weave Router Quota Drain Trap: Why Agencies Fail With Weave Router on Fixed-Fee Retainers
- Implementation BlueprintWeave Router Cost-Control Retainer Build (5-7 days)