Failure PatternDecision layer

The Bolt Token Burn Trap: Why Agencies Fail With Bolt on Fixed-Fee Client Builds

Symptom: Pro projects stall mid-build when the 10M monthly token allowance runs out and unused tokens from prior months are the only buffer left. Root cause: Bolt bills on token consumption, not seats or projects, so a single ambitious prompt session on a full-stack build can consume a large share of the 10M monthly Pro allowance with no warning threshold in the workflow.

By InnovaAI ResearchPublished

How do you recognize it?
  • •Pro projects stall mid-build when the 10M monthly token allowance runs out and unused tokens from prior months are the only buffer left
  • •Clients report Bolt branding still visible on a site the agency quoted as white-label delivery
  • •Team members on the Teams plan at $30 per member per month each burn tokens on the same client project because no per-project token cap exists
  • •Agency owners discover after the fact that the Free plan's 300K daily limit and 1M monthly ceiling were used for a paid client engagement
  • •Handoff drags because the client's own team cannot edit content without consuming tokens on prompts the agency already paid to run
Why does it happen?
  • •Bolt bills on token consumption, not seats or projects, so a single ambitious prompt session on a full-stack build can consume a large share of the 10M monthly Pro allowance with no warning threshold in the workflow
  • •The Free tier caps usage at 1M tokens per month and 300K daily while stamping Bolt branding on published sites, which makes it tempting for evaluation but unusable for client-facing delivery
  • •The Teams plan at $30 per member per month centralizes billing and admin controls but does not document multi-tenant client reporting, so agencies running many concurrent accounts cannot attribute spend per client
  • •Iterative prompting is the core build method, and re-prompting to correct generated code multiplies token draw in a way fixed-fee scoping rarely anticipates
How do you fix it?
  • •Move every client-facing build to Pro at $25 per month so the 100MB upload limit, custom domain support, and removal of Bolt branding apply before the first client sees a staging link
  • •Track token consumption per project weekly and set an internal rule that unused tokens rolling over from the prior month are reserved for revision rounds, not new builds
  • •Restrict Teams seats to the two or three people who actually prompt on a given account, since each member at $30 per member per month adds cost without adding shared token capacity
  • •Before quoting a fixed fee, run the client's core build once in a sandbox project and record the token draw, then price the retainer against that measured figure