Kimi Allegretto Plan
Date:August 27, 2026
Plan: Allegretto (US$ 39.90/month)
Subject: Request for immediate investigation, billing correction, and full refun d
1. EXECUTIVE SUMMA RY
I am an Allegretto subscriber and I am formally documenting a series of critical failures in the Kimi platform: abnormal credit consumption on simple tasks, undocumented feature restrictions, and non-existent customer support. These issues render the service defective and the pricing model decepti ve.
2. DOCUMENTED F ACTS
2.1 Abnormal credit consumption on simple poster revi sions
· On August 27, 2026, I requested visual revisions of a poster — layout adjustments, text alignment, minor tweaks. These are among the lightest possible tasks for any AI pla tform.
· The consumption occurred in two separate revision rounds for the same simpl e task:
· First revision: consumed
· 1.2% of monthly credits
· Second revision (same task, additional tweaks) : consumed
· 1.45% of monthl y credits
Total for poster revisions: 2.65% of monthly credits. No code generation, no long-document analysis, no multi-step reasoning, no complex image processing. Just visual adjustments to a poster.
Additionally, 3 messages in standard K3 mode (no Swarm, no Max mode) consumed over 2% of the monthly c redit pool.
2.2 Confirmed pattern from other users (o fficial forum)
This is not an isolated issue. On the official Moonshot AI forum (forum.moonshot.ai), other Allegretto subscribers reported iden tical behavior:
· A single K3 prompt in Kimi Code (standard mode, no Swarm) burned 100% of the 5h weekly limit and 20% of the monthly shared credit pool. Swarm activated automatically witho ut user consent.
· Skill creation (a local file operation with no API usage) consumed 5.09% of monthly credits in one go. This should not consume A PI tokens at all.
· Allegro subscribers (US$ 99/month) reported burning their entire monthly quota in just 2 to 9 day s of normal usage.
2.3 Arbitrary Skills restriction to K3 mode — not documented anywhere
· On August 27, 2026, when attempting to use a Skill in Kimi, the interface displayed th e following message:
“Skills are only available in Agent mode”(see attached screenshot)
This restriction is nowhere to be found in Moonshot’s o fficial documentation:
· The help center states Skills are supported in “Agent Mode” and “Kimi Claw” — with no mention of model -specific restrictions.
· The K2.6 product page explicitly states: “High-quality documents can now become reusable skills with Kimi K2.6” — confirming Ski lls should work on K2.6.
· There was no warning at signup that Skills w ould be locked behind K3.
Implications of this u ndocumented restriction:
Skills are a core feature that justifies the paid plan. Locking them arbitrarily behind K3 — the model with the cr itical consumption bug — is
· commercial gatekeeping
· that forces users into the most expensive mode t o access basic functionality.
· K3 is the exact model burning credits on simple prompts. Forcing users to use it for Skills mu ltiplies the financial damage.
· If K2.6 supports Skills (per official docs) but the UI blocks them, this is either a severe UX bug or a delibera te, uncommunicated restriction.
3. LACK OF TRANSPARENCY & HIDDEN LIMITS
When I subscribed to Allegretto at US$ 39.90/month, I expected usage volume at least 2x that of Claude Pro (US$ 20/mon th). The reality is the opposite:
· Moonshot does not disclose exact token/credit limits per operation at signup. Users only discove r the ceiling after being charged.
· There is no real-time per-prompt consumption dashboard — only aggregated percentages that prevent auditing.
· The 5h weekly Kimi Code limit is an additional hidden restriction not clearly stated during purchase.
· Now, Skills — a core paid feature — are arbitrarily restricted to K3 wit h zero prior notice or documentation.
In short: I paid for a “premium” plan that delivers less transparency, fewer features, and le ss control than cheaper competitors.
4. CO MPARISON: KIMI ALLEGRET TO vs CLAUDE PRO
Criteria
K imi Allegretto
Claude Pro
Monthly price
US$ 39 .90
US$ 20.00
Weekly li mit
5h (Kimi Code)
- share d monthly pool
None
Top model
K3 (critic al billing bugs)
Claude Opus (no weekly limit)
Transparency
Low — no per-prompt breakd own
High — visible, predictable usa ge
Support
Non-e xistent — email/WeChat ignored
Responsive — tickets in hours
Skill/tool creation
Consumes monthly credits (confirmed bug)
No extra charge
Auto-activati on of modes
Swarm t urned on without permission
Does not occur
Skills availability
Locked to K3 (undocumented)
Available in all modes
Conclusion: For half the price, Claude Pro offers a superior model, no weekly limits, functional support, no billing bugs, and unrestricted access to automation features. Allegretto costs twice as much, charges for operations that s hould be free, restricts features without warning, and off ers no effective support channel.
5. NON-EXISTENT SUPPORT — FORCED CONSUMPTION PRACTICE
Moonshot AI maintains support channels ([email protected], WeChat, for um) that do not respond. Multiple users report:
· Supp ort emails sent and ignored.
· WeChat w ith no human agent — only automated replies.
· Forum posts left unanswered for days.
When a company charges upfront, fails to deliver the contracted volume (due to billing bugs), refuses to communicate, offers no effective support, and continues automatic renewals, the practice resembles forced cons umption fraud-– or, at minimum, criminal negligence in billing management.
6. FORMAL DEMANDS
1. Immediate investigation of the billing bug that mult iplies credit consumption on simple operations (K3 standard, skill creation, auto-activated Swarm).
2. Public clarification on the Skills restriction: if it is a bug, fix it; if it is deliberate policy, document it transparently in the Terms of Service.
3. Correction of the token/credit counting system with transparent communication about what was changed.
4. Full refund of all Allegretto payments, given the service was d elivered in a defective state, with featu res restricted without notice, and with no support available.
7. REFERENCE — OFFICIAL FORUM THREAD
This complaint will be posted on the official Moonshot AI forum thread where the billing bug is already being discussed by the community:
I encourage all affected subscribers to contr ibute their own consumption data to this thread — the more evidence, the greater the pressure for a fix.
