# Fiitx

> Fiitx is a platform of AI employees — agents that do a job end to end inside the systems a company
> already uses: spreadsheets, Lark/Feishu, ERP, e-commerce back offices and chat. They follow the
> organization's own rules and stop for a human sign-off before anything leaves the building.

Fiitx is not a chatbot wrapper. Each AI employee is defined as a job: what it reads, which systems it
may touch, which numbers count as correct, and where it must ask a person first. The desktop app runs
on the employee's own macOS or Windows machine and holds the runtime, the browser sessions and the
credentials. External model calls may send prompts and task-relevant data excerpts to the selected
model service; the scope depends on the task and configuration. Enterprises can use their own
model or gateway. x.fiit.ai is the control plane for organization definitions, rules and entitlements.

## How it works

- **A job, not a prompt.** An AI employee owns a recurring piece of work — a month's attendance, a daily report, a round of creator outreach — and is judged on the output, not on the conversation.
- **Rules are delivered per organization.** Report formulas, approval thresholds, column meanings and tone are held in the cloud per organization and pushed to every seat. Correct a definition once and nobody repeats it.
- **Human sign-off on anything outbound.** Sending a message, posting content, paying, writing back to a system of record — each stops and shows the exact payload for approval before it goes.
- **Scheduled work.** Jobs run on a timer (overnight collection, an 8am report) without anyone opening the app.
- **Memory across sessions.** A corrected formula, a column name or a form of address is remembered and applied next time.

## AI employees (15)

### People, finance and operations

- [Attendance Analyst](https://fiit.ai/agents/attendance): Queries attendance, catches exceptions, builds the monthly report, alerts each department lead. Systems: Lark attendance, Lark messaging, Employee roster, Overtime and leave approvals, Excel / CSV.
- [Finance & P&L Analyst](https://fiit.ai/agents/finance): Reconciles several source sheets by your definitions, delivers P&L analysis by client, channel, project and company. Systems: Lark Base and sheets, ERP exports (account balances, dimensions), Offline Excel workpapers, Budget sheets, CRM (phased).
- [Performance Review Analyst](https://fiit.ai/agents/performance): Consolidates monthly review results; queries, compares and alerts by person, team, level and period. Systems: Lark Base, Lark Docs, Offline Excel review sheets, Employee roster.
- [Daily Report Builder](https://fiit.ai/agents/daily-report): Reaches through each department's source sheets, applies your formulas, posts the report on schedule. Systems: Lark spreadsheets, Lark messaging, Daily report base workbook, Department source sheets.
- [Creator Campaign Coordinator](https://fiit.ai/agents/creator-business): Creator intake, work orders, deadline chasing, client review sync, platform tagging — one pipeline end to end. Systems: Lark Base, Lark Docs, Lark messaging, Creator ad platforms, WeChat (connector).

### Cross-border e-commerce

- [Commerce Command Center](https://fiit.ai/agents/control-decision): Turns goals into standard tasks, routes them to the right agent, unifies KPI definitions, aggregates receipts and exceptions. Systems: Lark Base, Lark messaging, Shopee / TikTok Shop / Lazada connectors, Receipts from each business agent.
- [Shopee Operations](https://fiit.ai/agents/shopee-ops): Multi-shop product, stock, campaign and after-sales sync, a fixed-format daily report and exception alerts. Systems: Shopee Seller Centre / exports, Third-party ERP exports, Lark Base, Lark messaging.
- [TikTok Shop Operations](https://fiit.ai/agents/tiktok-shop-ops): Listing and campaign lifecycle, video and live publishing pipeline, traffic-to-conversion data and alerts. Systems: TikTok Shop Seller Center / exports, Lark Base, Lark messaging.
- [Creator Alliance & Outreach](https://fiit.ai/agents/creator-alliance): De-duplicates the creator pool, tiers and filters, batch-invites in the seller back office with your script, waits for your go-ahead, attributes results afterwards. Systems: TikTok Shop affiliate back office (Chrome extension), TikTok Shop Affiliate API (per-shop authorization), Lark Base, Lark messaging.
- [Product Selection](https://fiit.ai/agents/product-selection): Category heat, competitor price bands, multi-factor scoring with profit modelling, launch feasibility checks. Systems: Platform rankings / trend data, Lark Base, Supply-chain and logistics imports.
- [Creative Production](https://fiit.ai/agents/content-generation): Locks the product, writes copy and scripts, calls image / video / voice models to produce real assets, checks them, delivers. Systems: Image / video / TTS / subtitle models (Seedance, Jimeng, Kling, Hailuo, MiniMax and more), Asset library, Lark Drive, Platform creative spec library.
- [Multilingual Customer Service](https://fiit.ai/agents/customer-service): Classifies pre-sale / shipping / after-sales / refund / review intents, matches standard replies, drafts in Chinese, English and Thai, escalates risk. Systems: Ticketing / IM connectors, Standard replies and FAQ base, Lark messaging.

### General

- [Content Ops](https://fiit.ai/agents/content-ops): Decides what to post each day, writes from your marketing profile, has the Creative Production agent make the images and video, checks, publishes on time, then learns from the numbers. Systems: Marketing profile (per organization, self-service), Creative Production agent, WeChat Official Accounts, X, Mini-program analytics (optional), Public trending lists.
- [Viral Remake](https://fiit.ai/agents/viral-remake): Give it a video that performed. It breaks down the structure and cut rhythm, then remakes it with your own brand and selling points. Nothing is generated before you approve the shot list, and nothing ships before you approve the rough cut. Systems: Reference video (X link or uploaded file), Marketing profile (brand, products, selling points), Asset library (characters, products, brand fonts), Image, video and speech models, Local editing engine.
- [Custom Coworker](https://fiit.ai/agents/custom): Describe your SOP once. We turn it into a job description and deliver it to your workspace. Systems: Lark / WeCom / DingTalk, Base tables / spreadsheets / Excel, Email / calendar, Browser (real logged-in session), Local scripts.

## Use cases

Lines of work that have actually been delivered — where each one used to get stuck, how it runs once handed over, and where it stops for a human.

- [Month-end attendance: from three days of exports and chasing to one scheduled run](https://fiit.ai/use-cases/attendance-month-end): The first days of every month go into exporting attendance, reconciling what counts as overtime, and chasing each department. This is how that line runs once an AI employee owns it: your own pay period, approved-only overtime, per-department alerts, and a human check before anything goes out.
- [The company daily report: reaching through departmental source sheets, in the group chat before work](https://fiit.ai/use-cases/daily-business-report): The daily report is copied every morning out of a handful of departmental sheets, and the formulas live in one person's spreadsheet. This is how that line runs once an AI employee owns it: overnight collection, your own formulas, header changes detected rather than silently absorbed, and source sheets never written back to.
- [Creator outreach: turning 'scroll the back office, hope it is not a repeat, check before sending' into a traceable line](https://fiit.ai/use-cases/creator-outreach): Outreach lists are built by scrolling the seller back office, and a lot of what turns up has already been invited. Repeat invites damage the account, and the local ledger never matches the platform. This is how the line runs with an AI employee: three dedupe passes, a preview you approve, sending exactly that list, and reconciling afterwards.
- [Reconciliation and P&L analysis: from cross-checking sheets and asking around about definitions to one recomputable line](https://fiit.ai/use-cases/finance-reconciliation): Monthly analysis means joining several differently-shaped sheets across client, channel, project and company, while the definitions live in a few people's heads. This is how the line runs with an AI employee: columns matched by name rather than position, definitions delivered per organization, and results kept as snapshots you can recompute.
- [Multilingual customer service: from slow, inconsistent replies to one line with a standard](https://fiit.ai/use-cases/multilingual-customer-service): Cross-border support is rarely hard, it is messy: several languages, the same questions over and over, three agents answering three different ways, and the few tickets that genuinely need escalation buried underneath. This is how the line runs with an AI employee: classify first, one source of policy across languages, risky items escalated.
- [Monthly performance review: scoring, ranking, ABCD grading and the questions that follow, as one auditable line](https://fiit.ai/use-cases/performance-review): Every month three scores become a composite, the composite becomes an ABCD grade under a forced distribution, and then the questions start: who has two D's in a row, who is trending up. This is how the line runs once an AI employee owns it: company-issued definitions, a deterministic engine, the same answer every time, and name lists scoped to the asker's role.
- [Daily posts on a WeChat Official Account: topic, draft, images, compliance and publish as one scheduled line](https://fiit.ai/use-cases/official-account-daily-post): One promotional article a day: a topic that does not repeat, images at the right size, no wording the advertising law forbids, and a receipt from the Official Account backend proving it went out. This is how the line runs once an AI employee owns it: a fact sheet sets the wording, topics rotate by weekday, compliance runs before publishing, and every step is a real API call whose status is read back.
- [Viral video remakes: from "just make one like this" to a measured, approved process](https://fiit.ai/use-cases/viral-video-remake): A reference link isn't a brief. This page covers how an AI employee measures a hit video's real cut rhythm, rewrites it with your brand, and stops twice for your approval before anything ships.

## Guides

Definitions and comparisons — one question answered per page.

- [What is an AI employee?](https://fiit.ai/guides/what-is-an-ai-employee): An AI employee owns one recurring piece of work end to end inside the systems a company already uses, follows the organization's own rules, and stops for a person before anything goes out.
- [Fiitx vs. workflow automation tools](https://fiit.ai/guides/fiitx-vs-automation-tools): How an AI employee platform differs from trigger-action tools like Zapier, Make and n8n: what each is actually built to own, and where the two can work together.

## Verification examples

Synthetic fixtures processed by Fiitx's existing calculation modules. These are not client records, live platform demonstrations or measured time savings. Each page states prerequisites, outputs and limits and links to input JSON, result JSON and result CSV.
- [Attendance field mapping](https://fiit.ai/use-cases/attendance-month-end#example): leave-hour conversion and missing values versus explicit zero.
- [Financial consolidation](https://fiit.ai/use-cases/finance-reconciliation#example): subtotal exclusion, aggregate margin recalculation, source conflicts and missing metrics.
- [Creator invitation preview](https://fiit.ai/use-cases/creator-outreach#example): duplicate candidate removal and three-state verification with simulated platform responses. No real invitations are sent.

## Plans

- **Free** — 300 credits on sign-up, no card required.
- **Starter** — $20/month (CN ¥148), 5,000 credits/month; or $204/year (CN ¥1,510) paid yearly, saving 15%, with the full year of credits granted upfront. Chat, coding, local files and terminal; Standard image and video generation; Domestic model routes; 5 concurrent tasks · 5 scheduled tasks; Mac and Windows desktop.
- **Plus** — $40/month (CN ¥288), 10,000 credits/month; or $408/year (CN ¥2,938) paid yearly, saving 15%, with the full year of credits granted upfront. Everything in Starter; Premium model tier for chat, image and video; Domestic and overseas routes; Larger single jobs; 20 concurrent tasks · 20 scheduled tasks.
- **Pro** — $200/month (CN ¥1,448), 50,000 credits/month; or $2,040/year (CN ¥14,770) paid yearly, saving 15%, with the full year of credits granted upfront. Everything in Plus; 5x the usage of Plus; Priority at peak hours; Batch and long-running jobs · 99 scheduled tasks; Priority support.

Credits meter model usage. Text work costs a few credits per turn; image and video generation cost more because the upstream models do.

## Languages

- [English](https://fiit.ai/)
- [简体中文](https://fiit.ai/zh)
- [繁體中文](https://fiit.ai/zh-TW)
- [日本語](https://fiit.ai/ja)

## What Fiitx is not

- Not a hosted SaaS that ingests your data: the runtime is the desktop app on your own machine.
- Not an autonomous agent that acts unsupervised: outbound actions require a person to approve them.
- Not a model provider: Fiitx routes to third-party models, and an organization may bring its own API keys.

## Links

- [All AI employees](https://fiit.ai/agents): the full public catalog
- [Use cases](https://fiit.ai/use-cases): how a line of work runs once handed over
- [Guides](https://fiit.ai/guides): definitions and comparisons
- [Download for macOS](https://fiit.ai/api/download/mac)
- [Download for Windows](https://fiit.ai/api/download/windows)
- [Sign in / create an account](https://fiit.ai/login)
- [Privacy policy](https://fiit.ai/privacy)
- [Terms](https://fiit.ai/terms)