v1.5.0 to v1.7.0

---

name: senpi-account-status

description: >-

does Senpi cost?", "what do I get on Pro/Quant?", "how do I get free credits?", "where are my

milestone credits?", "do I need a subscription to keep my strategy running?", "what happens when my

license: Apache-2.0

metadata:

author: Senpi

platform: senpi

exchange: hyperliquid

---

# Senpi Account Status — your standing across Senpi

A hidden engine pulls the user's standing in one real-time call; **your job is to present it

## Golden rules

referral together. Read its JSON.

- **Real-time, not from memory.** Points/rank move — never quote a figure from earlier in the

conversation; re-run.

- **Credits are explained, never read.** No tool returns the AI-credit balance; the meter in the app

header is the balance. Rules in "AI credits — the usage meter" below.

## How to run the engine

```

python3 scripts/status.py

```

- `referral` — `balance_usdc` (pending referral earnings, 25% of builder fee on referred trades).

- `meta.warnings` / `meta.degraded` — narrate honestly.

- Fails open — partial data still returns valid JSON.

## Output contract

Lead with the section they asked about; otherwise a tight standing card:

1. **Points & rank** — total, base/perp split, multiplier, rank (+ change).

3. **Referral** — pending `balance_usdc` (and that it's claimable if > 0).

## AI credits — the usage meter

Senpi agents run on **AI-usage credits that come with the plan** — a trial grant first, then plan

credits. The app shows what is left as the **meter in the app header** (users call it "the bubble" or

"my credits"), and their subscription page is the source of truth for the number. What each plan

includes is published in "Plans & free credits" below and may be quoted; how much is **left** depends

on your plan and your usage, and only the user's own page shows that.

- **What spends them.** Every message the user sends and every tool turn the agent runs. Long threads

cost more because the whole conversation is re-sent each turn — a fresh thread is cheaper than a

long one. Scheduled monitoring (a cron that re-reads a strategy) is the classic silent drain;

`senpi-strategy-author/references/shadow-testing.md` has the cheap alternatives.

- **A running strategy is not what spends them.** Its runtime ticks without a model call, so running

it does not move the meter; a `decision_mode: llm` action in its `runtime.yaml` is the exception,

since that calls a model when it fires. A cron the agent set up to check on a strategy does move it

(each firing is a full model call), so name any the agent runs for this user. Using no credits is

not the same as needing no plan: the strategy runs on the agent, and a lapsed plan pauses both

(see "Plans & free credits").

- **Not trading money.** Credits cannot be traded, withdrawn, deposited into a strategy, or moved

anywhere — they are separate from the funding wallet and from every strategy budget. A shrinking

meter is usage, never a trading loss; a losing trade never touches the meter.

- **You cannot read the balance.** No Senpi tool returns it (`user_get_me` → id / wallets / createdAt

section. Point at the meter in the app header, or at their subscription page for the exact figure —

the in-app indicator surfaces only when credits run low, so a missing indicator is expected, not a

lost balance. Never invent a number, a percentage or a "roughly", and never guess what the meter

shows.

- **Chat is not free of credits.** Every turn spends them — including the turn that asks what a turn

cost. Say so plainly; never "nothing".

| Say | Never say |

|---|---|

| "Chat and tool use spend AI credits from your plan; the meter in the app header is the balance — I can't read it from here. Longer threads cost more because the whole conversation is re-sent each turn." | "chatting is free" / "that cost nothing" |

| "What each plan includes is published — here's the table. How much is left depends on your plan and your usage; the meter shows that." | Any balance or percentage you did not read from the app, or a plan size you did not read from the table below |

| "Credits aren't money: they can't be traded, withdrawn or put into a strategy — your funding wallet and strategy budgets are separate." | "credits" as a tradable or withdrawable balance |

| "The meter going down is usage — it moves with messages and tool turns, not with your trades." | Blaming a shrinking meter on trading losses |

| "Your strategy runs on its own without the AI, so running it doesn't use credits. Chat does, and so does any scheduled check-in I run for you." | "running the strategy costs nothing" while the agent runs a check-in cron on it; a wallet or strategy balance offered as proof about credits |

| "Your strategy doesn't use credits, but it runs on your agent, and the agent needs an active plan. If the trial ends, a renewal fails or the credits run out, the agent and your strategies pause until you subscribe or credits return." | "no subscription is needed" / "your strategy keeps running without a plan" |

**Routing.** "What is the bubble / meter in the header?", "why are my credits going down?", "how much

did that cost?", "does my strategy use credits?", "can I trade / withdraw / deposit my credits?" all

come here, in any language. "Credits" (Ukrainian "кредити", Spanish "créditos") means these AI credits

unless the user says leverage, a loan or funding; when it is unclear, ask which before answering. A

credits question is answered from this section alone — no engine run, no number — and closes by

does it cost / what do I get on Pro / how do I get free credits / where are my milestone credits /

do I need a subscription / what happens when my trial ends" are **catalog** questions: answer them from "Plans & free credits" below, still without a figure

about this user.

## Plans & free credits

Published facts — quote them. Billed per **30-day cycle, rolling from the user's own start date**,

never a calendar month.

| Plan | Price | Credits | Points per $1 volume |

|---|---|---|---|

| Starter | $25 | $25 | 1 |

- **Every plan includes every model.** Plans differ on credits and points rate only — never on model

- **$1 of credit = $1 of AI usage.** Full credits land at the start of the cycle and expire at the end

of it: **no rollover, no top-ups, no overage**. Out of credits = upgrade (instant) or wait for renewal.

- **Upgrading is instant, any day** — the new plan's full credits land immediately, the 30-day cycle

restarts, and unused credits are credited against the price, so they pay the difference.

**Downgrading or cancelling takes effect at the next renewal**, never mid-cycle; benefits are kept

until then and there are no mid-cycle refunds.

- **Payment is web only** (Stripe, senpi.ai/settings/subscription). The app shows plan, credits and

renewal date but links out to the web page to change anything.

- **Free trial:** activates when the user **deploys their agent**, is one-time, and expires **14 days

after activation**. **Never quote a trial figure** — the starting amount has changed more than once;

their subscription page is the only source of truth. On expiry or exhaustion the agent stops until

they subscribe: active strategies pause, **open positions are NOT auto-closed**, funds untouched.

- **A lapsed plan pauses everything, not just chat.** Strategies run on the agent. When a trial or

plan expires, a renewal payment fails, or the cycle's credits run out, Senpi pauses the agent and

every strategy on it stops with it: no scans, no new entries, no stop moves. Open positions stay

open, and a stop-loss already resting on the exchange still triggers. Funds stay in their wallets,

withdrawable in the app. Subscribing, a successful renewal or an upgrade brings the agent and its

strategies back. You cannot see the user's plan, trial or renewal dates: point at

senpi.ai/settings/subscription, and never tell them a subscription is not needed.

## ⚠ Token scope

Every tool here is **USER-scoped** (the user's own account): needs a USER-scoped `SENPI_AUTH_TOKEN`.

App-scoped → no user resolves and `meta.degraded`; say so rather than reporting zeros.

## Skill Attribution

Guide/analysis skill — it *reads* the user's standing; it does not mutate anything (claiming referral

rewards is a separate explicit action via `user_claim_referral_rewards`, which this skill never calls).

## Install — both scripts are required

The engine is **two files** in `scripts/`: `status.py` (the engine) and `mcp_client.py` (its vendored

MCP helper, imported at runtime). **Install the whole `scripts/` directory** — copying `status.py`

alone fails with `No module named 'mcp_client'`. Stdlib only, no other runtime dependencies.