Pricing
A diagnosis is one completed plain-English root cause — the moment Mushi Mushi finishes reading a report and can hand your editor a paste-ready fix prompt. Noise filtered out or duplicate reports collapsed to one row do not count. You pay for comprehension, not volume.Start free — 1 app, 100 diagnoses/mo, no cardMove to Builder for a second app. Upgrade from Billing, no sales call.
Plans
| Plan | Apps | Monthly | Annual (2 mo free) | Diagnoses | Retention | Seats | What you get | Get started |
|---|---|---|---|---|---|---|---|---|
| Self-host | Unlimited | Free forever | — | Unlimited (your own AI key) | Your choice | Unlimited | Full OSS stack — MIT SDK, AGPLv3 server, bring your own LLM keys | Read the guide |
| Free | 1 | $0 | — | 100 / mo | 14 days | 1 | One app, the whole loop once: plain-English diagnosis and a fix packet in your editor; agent fix PRs once you connect GitHub. Hard stop at the limit | Start free |
| Builder | 3 | $29 / mo | $290 / yr ($24.17/mo) | 1,000 / mo | 90 days | 1 | Three apps, plugins from the marketplace, usage alerts and an audit log; then $0.03/diagnosis with a $50 overage cap you set | Start free, upgrade later |
| Studio | 10 | $79 / mo | $790 / yr ($65.83/mo) | 3,000 / mo | 180 days | 5 | A portfolio of apps: 5 seats, the weekly report across your apps, longer history; then $0.03/diagnosis up to a $200 cap | Start free, upgrade later |
| Enterprise | Unlimited | From $499 / mo | Custom | Custom | Up to 365 days | Unlimited | SAML SSO, SOC 2 evidence collection (not a certification), priority support | From $499/mo — contact us |
A plan is the number of apps Mushi Mushi watches for your team: one on Free, three on Builder, ten on Studio. Diagnoses are the allowance under it. Reports are always captured and are never billed.
Installing plugins from the Marketplace (PagerDuty, Zapier, the Cursor and Claude Code agents, Linear Sync, Sentry Mirror) needs Builder or above. The Integrations page (Sentry, Slack, GitHub, Linear) works on every plan.
Annual billing saves two months. Pay for 10 months, get 12: Builder is $290 a year, Studio $790. Annual plans have no pay-as-you-go overage: they include the plan’s diagnoses each month and pause new diagnoses at that limit. Pick annual at checkout, or switch an existing plan with Change plan on the Billing page. For Enterprise commits, contact us.
Overage is capped. Builder and Studio include a default cap on overage spend ($50 and $200 a month, on top of the base fee). You can lower or raise it in the console. At the cap, Mushi Mushi pauses new diagnoses and says so, instead of sending a bigger invoice. Email alerts fire at 50%, 80%, and 100% of your included diagnoses.
Free has a hard stop, not a soft limit. At 100 diagnoses Mushi Mushi stops diagnosing new reports until the next calendar month. No card, no charge, no bill shock.
What counts as a diagnosis
One diagnosis = one completed classification pass. Specifically:
- The report passes the fast noise filter (not spam or incomplete junk).
- It is not a duplicate of an existing group already classified this period.
- The classifier finishes and writes a plain-English root cause + severity + fix hint.
Under the hood that is a two-stage pipeline (Haiku fast-filter → Sonnet structured output). Classification, deduplication, knowledge-graph writes, and MCP tool calls are not counted separately: they are included in the per-diagnosis price. Fix PRs and agent runs are not counted as diagnoses; with your own AI key (see below), their tokens are billed by that vendor.
What does NOT count:
- Reports filtered by the fast-filter (noise, spam, incomplete submissions).
- Reports deduplicated to an existing group (same broken button, same stack trace: they collapse, you only pay once).
- Inbound adapter events (Sentry alerts, Datadog webhooks) that don’t complete a fresh diagnosis.
- MCP
get_fix_contextcalls, fix dispatch, or any read-only operation.
Pricing estimator
Use the slider below to model your bill. Toggle Annual to see the effective monthly base fee. When you hit a tier limit, the estimator suggests the next plan.
Interactive pricing estimator
Estimates use list pricing. Monthly plans meter overage up to the spend cap; annual plans save 2 months on base fees and stop at the included diagnoses. Self-host for free using your own LLM key.
Example bills
Monthly plans, default caps, for a billing period that lines up with the calendar month (overage metered on top of the base fee; the cap limits the overage, not the total):
Builder: $29 + min(max(0, diagnoses − 1,000) × $0.03, $50)
Studio: $79 + min(max(0, diagnoses − 3,000) × $0.03, $200)| Diagnoses / mo | Builder | Studio |
|---|---|---|
| 100 | $29.00 | $79.00 |
| 500 | $29.00 | $79.00 |
| 1,000 | $29.00 | $79.00 |
| 1,500 | $44.00 | $79.00 |
| 2,000 | $59.00 | $79.00 |
| 3,000 | $79.00 (cap) | $79.00 |
| 5,000 | $79.00 (cap) | $139.00 |
| 8,000 | $79.00 (cap) | $229.00 |
The cap is checked per calendar month (UTC) of usage: at the cap, new diagnoses pause until the 1st. Builder reaches its default cap at about 2,667 diagnoses in a month, Studio at about 9,667. Reports past that point are kept, not diagnosed.
Why apps, not reports?
Reports are noisy: one broken button can send fifty, and a busy week is hard to predict. The number of apps you ship is something you already know, and it grows with what you get out of Mushi Mushi. So a plan is sized by apps watched, diagnoses are a generous allowance under it, and reports are never billed at all.
Why diagnoses, not events
Sentry meters errors. We meter diagnoses: one per plain-English root cause and paste-ready fix, a unit you can predict. Duplicate errors from the same broken button collapse to one diagnosis. You pay once.
Your own AI key and the self-host path
On every Cloud plan, Free included, you can add your own Anthropic, OpenAI or Cursor key under Settings → API keys. Mushi Mushi then uses it for fix PRs and agent work; you pay those vendors directly at list rate, and we never mark up tokens.
When self-hosting, Mushi Mushi uses your keys for everything. For a sense of scale: on Mushi Mushi’s own key, a diagnosis cost about $0.017–0.022 in LLM tokens over the 30 days to 2026-10-28, so 1,000 diagnoses is roughly $20. Fix PRs and reviews are extra and depend on the model you pick.
The self-host path is free, and always will be: see the guide.
Compared to Sentry + Seer
| Sentry Team + Seer | Mushi Mushi Builder | |
|---|---|---|
| Base plan | Team: $26 / mo billed annually, $29 monthly | $29 / mo for 3 apps ($290 a year) |
| AI diagnosis | Seer add-on: + $40 / active contributor / mo | Included |
| Solo dev total | $66–69 / mo plus any pay-as-you-go | $29 / mo |
| Catches non-crash friction | Via its User Feedback widget and Session Replay | Yes: a report widget (shake on mobile) with screenshot and console/network tail, then a diagnosis |
| Bugs in your editor | Sentry MCP server: issues and Seer analysis in Cursor / Claude Code | Mushi Mushi MCP server: diagnosis, fix prompt and lessons; no second LLM key |
Sentry prices from sentry.io/pricing and Sentry’s pricing docs , checked 2026-10-29. Seer is an add-on to the Team or Business plan, billed as its own monthly charge. An active contributor is anyone who opens two or more pull requests in a month to a Seer-enabled repo; the $66–69 total assumes one of them. Seer is not available in self-hosted Sentry (self-hosted docs ).
Sentry + Seer is strong at code-thrown errors, tracing, breadcrumbs and session replay, and Seer is an AI debugging agent for those issues. Mushi Mushi starts from user-felt friction that never triggers an error, ingests Sentry’s errors too, and turns each into a plain-English diagnosis and a paste-ready fix prompt in your editor. A diagnosis is an LLM’s reading and can be wrong; check it before you merge.
Frequently asked questions
What happens when I hit my diagnosis limit?
Free: New plain-English diagnoses pause until the next calendar month. Reports are still captured.
Builder / Studio (monthly): You can keep diagnosing into overage until your spend cap stops new classification runs. Reports are never dropped.
Builder / Studio (annual): New diagnoses pause at the included amount until the next calendar month. Reports are never dropped. Switch to monthly (pay-as-you-go overage) or to Studio from Change plan.
What happens when I add one app too many?
The console says how many apps your plan watches and how many your team has, and offers the plans. Nothing already running stops: your existing apps keep their reports and diagnoses. Move up a plan, or remove an app you no longer use.
Can I use my own AI key on Free?
Yes. Bring-your-own-key works on every plan, Free included. Add an Anthropic, OpenAI or Cursor key under Settings → API keys and Mushi Mushi can open fix PRs with it, so you see the whole loop on your first app before paying anything.
When does my quota reset?
Diagnoses are counted per calendar month (UTC) on every plan, so the count starts again on the 1st. Paid plans are invoiced on your Stripe billing period, which can start on a different day.
Can I change my spend cap?
Yes. Open Billing → Spending cap on your project card. Set any USD cap on overage from $0–100,000 or clear it to use the plan default ($50 Builder, $200 Studio).
Who receives usage alert emails?
By default, the project owner. Override the address under Usage alert email on the same card. Alerts fire once per threshold (50%, 80%, 100%) per billing month.
Is usage per app or per team?
Diagnoses are per app; the app limit is per team. Each app you connect has its own diagnosis counter and spend cap, and the plan’s app limit counts the apps in your team.
Can I downgrade mid-cycle?
Change plans in the console with Change plan on the Billing page (Builder ↔ Studio, monthly ↔ annual). The change applies immediately and is pro-rated: unused time is credited and the new plan is charged for the rest of the period. Moving between monthly and annual starts a new billing period that day. Stripe’s customer portal (Manage) handles your card, invoices and cancellation.
Do annual plans include overage?
No. Annual plans include the plan’s diagnoses every month (1,000 on Builder, 3,000 on Studio) and pause new diagnoses at that limit until the next month, so an annual invoice never grows. If you need more, switch to monthly for pay-as-you-go overage or upgrade to Studio from Change plan.
What does Enterprise cost?
From $499 a month, billed annually: unlimited apps, SAML SSO, the SOC 2 readiness pack, self-host support and up to 365 days of history. [Contact us](mailto:kensaurus@gmail.com?subject=Mushi Mushi%20Enterprise%20inquiry) for a quote.
How do I estimate before signing up?
Use the interactive estimator above, or run mushi billing status after wiring the SDK: it shows diagnoses used, limit, and spend cap for the current period.
Start free
Start free — 1 app, 100 diagnoses/mo, no cardMove to Builder for a second app. Upgrade from Billing, no sales call.
Prefer the terminal? npx mushi-mushi runs the wizard: detect framework, install the SDK, write env vars, connect MCP. See the incident loop quickstart, or create your free account in the console .