If your team already pays for a coding agent, why should its feedback tool sell you another pile of tokens? Quell captures the evidence, applies the trust rules and serves structured context to the agent you already use. No token tax. The AI is yours; the feedback rail is ours.
What is a token tax?
A token tax is the second meter that appears when a SaaS tool owns model calls on your behalf. In feedback software it often sits behind AI triage or auto-categorisation, so the bill moves with submission volume and agent activity rather than team size.
This is not an argument that inference should be free. It is an argument about the product boundary. Capturing a failed request, preserving a console error and enforcing a permission rule do not require Quell to own another model call.
The problem with metered AI in feedback tools
Three issues compound each other:
- You pay twice — once for the feedback tool, once for the AI tokens it burns on your behalf. You cannot choose a cheaper model or run nothing when the AI adds no value.
- The AI is a black box. You cannot inspect the prompt, tune the classification logic, or swap the model for a newer one that outperforms the default.
- Agent workflows break the model. If your coding agent is reading and writing feedback continuously, per-event AI costs scale with agent activity — not with team size.
Teams that ship with Claude Code or Cursor already have an AI budget. The last thing they need is a feedback platform consuming a second, unpredictable slice of it.
How Quell avoids it
Quell has no built-in AI. The platform captures and structures feedback — screenshot, console logs, network calls, page state. It stores that data and exposes it via a REST API, a CLI, and an MCP server. What the agent does with the data is entirely up to you.
# fetch full context for agent triage quell context 412
The MCP server gives the agent 10 tools. None of them call an external AI API. The agent brings the intelligence; Quell brings the plumbing. The absence of a bundled model is not the missing feature. It is the boundary we chose.
Flat pricing model
Quell charges per seat, not per event. Managed cloud is £3/$4/€4 per user per month (early access now — no self-serve checkout yet) regardless of how many feedback items you receive, how often your agents poll the API, or how many issues get auto-routed to GitHub. There is no token meter, no AI credit top-up, no usage spike. Self-host is on the roadmap; the Dockerfile is in the public repo.
Honest tradeoffs
Bringing your own AI means you need an agent to do the triage. If your team does not use Claude Code, Cursor, or a similar coding agent today, Quell's MCP-based workflow is not a fit — you would be wiring up a new tool just to use the automation layer. Quell is honest about this: the product is built for teams who already have an agent in their loop.
For those teams, the no-token-tax constraint is a feature, not a limitation. You choose the model, you control the cost, and you keep the feedback loop deterministic.