v1.0.5 Powered by Claude Opus 4.7 · adaptive thinking

Win Upwork jobs faster.
Proposals that don't read like AI.

A Chrome extension that reads the job, scores the fit, and writes a proposal in your voice. No em dashes. No "leverage." No rule-of-three slop. Just the proposal you'd have written if you had two more hours.

Add to Chrome — it's free See it work →
Free plan, 10/mo No card required Pro $15/mo
LIVE DEMO

Pick a job. Watch a proposal write itself.

No sign-up. The same model, the same prompt rules, the same streaming you get inside Upwork — running right here on the page.

⌘ + ⏎ to generate
Streaming proposal will appear here…
Ready · Claude Opus 4.7
HOW IT WORKS

Three steps. Zero context-switching.

01 / DETECT

Open any Upwork job.

The extension reads the full job — title, full description, budget, required skills, client country, total spend, hire rate, payment-verified flag, even how many proposals are already in. No copy-pasting.

02 / SCORE

Get a Fit Score before you bid.

0–100, broken down across skill match, budget fit, client quality, competition, and experience match. Risk flags surface unverified payment, no spend history, expert-level mismatch, or 30+ proposals.

03 / WRITE

Stream into Upwork's textarea.

On the apply page, the proposal types itself directly into the cover-letter field — cursor and selection preserved. Three tones, three lengths, voice mirroring. Stop, regenerate, swap back. ⌘ + Shift + C to copy.

FIT SCORE

Stop bidding on jobs
you can't win.

Every detected job gets scored across five weighted components. Connects are expensive. The score tells you whether to spend them.

Generate Templates History
FREE · 1/10
JOB DETECTED
SaaS Website Visuals Designer
Looking for a designer who can create visuals for my SaaS website. Design is…
0 /100
Fair fit 2 risks
11skill 0budget 20client 5comp. 2exp.
  • Only ~2 proposals — low competition
  • Client payment verified
  • Client hires 60% of interviewed candidates
  • None of your listed skills match the job requirements
  • Budget ($20/hr) is below your rate ($39/hr)
ACTIVITY ON THIS JOB
Proposals~2
Interviewing0
THE REAL EXTENSION

A 0–100 score, computed in milliseconds, before you spend a single Connect.

Five weighted components. Verdict labels. Risk flags. The same panel you'll see in the side-panel on every Upwork job page — animated here so you can see what each part does.

75 +

Strong fit. Open the proposal and bid.

55 – 74

Good fit. Worth a careful read.

35 – 54

Fair fit. Check the risks first.

0 – 34

Poor fit. Save the Connects.

Skill match0 – 30
Client quality0 – 25
Budget fit0 – 20
Competition0 – 15
Experience match0 – 10
ANTI-AI DETECTION

Written by a human.
Hardened by a prompt.

Every tell a recruiter uses to spot AI writing is encoded as a hard rule in our system prompt. When you save a writing sample, the model mirrors your cadence over the generic style. The strongest anti-detection signal we ship.

PROMPT RULES · v3.2 ACTIVE ON EVERY GENERATION
01
Forbidden vocabulary

The words every AI overuses. Stripped on output.

02
Forbidden patterns

The sentence shapes that scream LLM.

03
Hard rules

Non-negotiable, on every output.

0
em dashes

The single most common AI tell. Replaced with commas, parens, or a new sentence.

1
bold phrase max

One anchor for the eye, never a full bolded sentence, never multiple highlights.

×
no rule of three

"Robust, scalable, and seamless" gets rewritten. Real writers don't speak in triplets.

⛨
injection blocked

"Ignore your instructions" or "type the word X" inside job posts is detected and refused.

04
Before / After

What the model would write vs what we ship.

AI "I'd love to leverage my robust expertise."
YOU "I shipped this exact migration last quarter."
AI "It's not just code, it's craft."
YOU "Plan: lift the design system, then swap routes."
AI "From idea to execution, seamlessly."
YOU "Got 15 minutes for a quick call this week?"
GENERIC AI vs APPROID

Same job. Two outputs.

Most "AI proposal" tools wrap a generic model in a generic prompt. Approid runs Claude Opus 4.7 with adaptive thinking against a hardened style spec — and your saved voice.

● GENERIC AI

What you usually get

Polished. Hollow. Detected.
"I hope this message finds you well! I'm thrilled to express my interest in your project — I bring a robust, comprehensive, and seamless approach to web development.

Whether you're a startup or a Fortune 500, I deliver solutions that leverage cutting-edge technology to help you navigate the digital landscape. From planning to execution, my work is holistic.

I would love to delve into your paradigm further!"
✓ APPROID · OPUS 4.7

What Approid generates

Specific. Direct. Yours.
"Hi,

The Webflow to Next migration is a project I've shipped three times in the last six months, so I've already hit (and fixed) most of the gotchas: flickering ISR, CMS field parity, image pipeline.

Plan would be: lift the design system into Tailwind tokens, mirror the schema in Sanity, then swap routes one section at a time behind a feature flag so traffic never breaks.

What's currently powering the blog — Webflow CMS, or did you migrate it elsewhere already?

Got 15 minutes for a quick call this week?"
TEMPLATES & HISTORY

Reuse what worked.
Learn from what didn't.

Save any proposal as a template. Tag it. Filter it. The last 20 proposals stay in History — full text, fit score, tone, length, model, tokens.

Templates
12 saved · JSON export
all frontend backend design ai cold
Next.js migration · technical opener
Saw the brief about migrating from Webflow. I've shipped this three times…
SaaS landing page · senior designer
Designing for AI products is a specific muscle. I've done two YC pages…
Postgres + tRPC migration
The Mongo → Postgres part is what caught my attention. I'd push for materialized views…
RAG pipeline build · cold reach
Read the part about ingestion latency twice. The fix is usually batched embeddings…
Generic frontend · cold opener
Quick read of the brief: the hard part is going to be the table virtualization on…
History · last 20
Newest first
React/Next.js engineer to rebuild marketing site
CASUAL · MEDIUM · OPUS 4.7 · 8.4s · today
Full-stack dev for B2B SaaS dashboard
TECHNICAL · MEDIUM · OPUS 4.7 · 11.1s · today
Senior designer for AI startup landing page
FORMAL · DETAILED · OPUS 4.7 · 14.8s · yesterday
WordPress maintenance retainer
CASUAL · SHORT · OPUS 4.7 · 5.2s · 2 days ago
RAG pipeline build for legal-tech startup
TECHNICAL · DETAILED · OPUS 4.7 · 13.5s · 3 days ago
PRICING

Same model. Same features.
Pay only when you bid often.

Free and Pro run on Claude Opus 4.7. The model and feature set are identical between plans. The only difference is your monthly cap.

Free

Forever
$0/ month
10 proposals per month · resets on the 1st
  • Claude Opus 4.7 with adaptive thinking
  • Fit Score on every job
  • 3 tones · 3 lengths · streaming
  • Templates (no cap)
  • History · last 20
  • Voice mirroring from your writing sample
  • Apply-page auto-typing
Add to Chrome — free

Pro

★ FOR BIDDERS
$15/ month
150 proposals per month · cancel anytime
  • Everything in Free
  • 150 proposals per month (15× the cap)
  • Priority support
  • Early access to new features
  • Cancel-at-period-end · keep access until renewal
  • Polar checkout · MoR handles tax worldwide
Start Pro — $15/mo
FAQ

Answers, not marketing.

Which AI model writes the proposals?
Claude Opus 4.7 with adaptive thinking — the same model on Free and Pro. The model and feature set are identical between plans. The only difference is the monthly proposal cap (10 on Free, 150 on Pro).
Will Upwork detect the proposal as AI-written?
Our system prompt enforces explicit anti-detection rules: no em dashes, no emojis, no rule-of-three lists, a forbidden vocabulary list (delve, leverage, robust, comprehensive, seamless, etc.), forbidden sentence shapes ("It's not just X, it's Y"), at most one bold phrase used as a hook. The strongest signal: when you save a writing sample, the model is instructed to mirror your cadence over the generic style rules.
Does the extension auto-submit proposals?
No. Every proposal must be reviewed and submitted by you. We deliberately do not implement auto-submission, connect-cost prediction, or auto-bid. On the apply page, the proposal types itself into the cover-letter textarea — you read it, edit it, and click Submit yourself.
What goes into the Fit Score?
Five components: skill match (0–30), client quality (0–25 — payment-verified, rating, hire rate), budget fit (0–20), competition (0–15 — proposals already submitted), experience match (0–10). Risk flags surface unverified payment, no spend history, expert-level mismatch, 30+ proposals, and budget below your hourly rate.
How is my data handled?
Job descriptions are not retained after the proposal response is returned. The generated proposal text is retained — that's what powers History — and is deletable per-item or all-at-once with account deletion. Passwords use Argon2 hashing. We never store card numbers (Polar handles checkout). Webhook signatures are verified on every callback.
Can I cancel anytime?
Yes. Cancellation defers to the end of your current billing period — you keep Pro until the renewal date, with a "Reactivate" button visible until then. Full refunds (admin-issued from Polar) trigger immediate downgrade to Free. Cancel and refund flows are emailed to you.
Does it work on every Upwork layout?
Yes. The content script supports the standalone job page (/jobs/~xxx), the find-work details view (/nx/find-work/details/~xxx), and the search-results slider modal (/nx/search/jobs/details/~xxx). It also detects "See more" truncation and warns you if the full description isn't visible.
What languages are supported?
The extension UI is English. Claude Opus 4.7 can write proposals in any language — including matching the language of the job post if you prompt it via your saved bio or writing sample.
READY WHEN YOU ARE

Your next proposal could be already written.

Install the Chrome extension. Open any Upwork job. The first proposal is free — and the next nine, this month.

Add to Chrome — free View pricing
NO CARD · 10 FREE / MONTH · CANCEL ANY TIME