finch-cli 0.3.0

Tailor your resume to any job posting from your terminal.

finch-cli

a terminal client for job hunting. browse fresh internship + new-grad postings, tailor your resume to any of them, see an ats-style match score with matched + missing keywords, keep a library of the tailored versions.

install

pip install finch-cli
finch login
finch ui

finch login opens a sign-in link in your browser. once you sign in with google or email on applyfinch.com, the cli is paired and tailoring just works -- no api key to manage, the backend pays the llm bill.

prefer your own key? skip finch login and set one instead:

export DEEPSEEK_API_KEY=sk-...
finch ui

deepseek keys are cheap (~30x cheaper than the major frontier models). get one at https://platform.deepseek.com/api_keys. any openai-compatible chat-completions endpoint also works (openai, together, groq, fireworks): pass --base-url and --api-key.

what it does

three tabs, one keyboard.

jobs pulls the simplifyjobs internship and new-grad lists (the same two feeds the finch backend mirrors). 3000+ active postings, filterable by query, source, term. press T on a row, the job loads into the tailor tab pre-filled.

library is the list of tailored resumes you've saved. Markdown preview on the right.

tailor is the three-pane editor: base resume / job posting / tailored output. the model rewrites bullets and ordering to fit the posting. won't invent skills you don't have. a match-analysis panel at the bottom shows an ats-style score, which job keywords your resume covers, which ones it doesn't, and how much the tailored version moved the needle vs the base.

install (pure cli, no tui)

finch tailor --resume resume.md --job https://jobs.example.com/swe-intern -o tailored.md

if the job site is js-heavy (workday, some greenhouse iframes), the fetch fails clean and tells you to use --job-file with a pasted description:

finch tailor -r resume.md --job-file pasted.txt -o tailored.md

tui keys

key	action
1 / 2 / 3	jump to jobs / library / tailor
Ctrl+T	tailor; on jobs tab loads the selected job first
Ctrl+U	paste a job url, fetches it into the tailor pane
Ctrl+O	open a resume file
Ctrl+R	refetch the job feeds
Ctrl+L	save the tailored output to the library
Ctrl+S	save the tailored output to a file
Ctrl+D	load the bundled demo
Ctrl+Q	quit

filter inputs at the top of the jobs tab are live; type and the table updates.

cli flags

finch tailor

flag	what
-r, --resume PATH	base resume in markdown or plain text (required)
-j, --job URL	job posting url
--job-file PATH	local file with the job description (use if --job won't extract)
-o, --out PATH	save tailored resume here, else stdout
--model NAME	model name (default deepseek-chat)
--api-key KEY	override $DEEPSEEK_API_KEY / $FINCH_API_KEY / $OPENAI_API_KEY
--base-url URL	override openai-compatible base url (default https://api.deepseek.com)

finch ui

flag	what
--demo	preload the bundled example into the tailor tab
--model NAME	model name

where stuff lives

	path
job feed cache (~6h ttl)	~/.cache/finch-cli/{intern,newgrad}.json
saved tailored resumes	~/.local/share/finch-cli/resumes/
bundled example resume + job	examples/ in the repo

why a cli

most of these tools are saas you have to sign up for. this is the opposite. install, run, the only thing leaving your machine is the llm call you control.

the autofill side - the part that actually fills out the application form - lives in the web app at applyfinch.com, free during beta.

known limits

• workday and some greenhouse iframe pages need js. the fetch fails clean and tells you to use --job-file.
• tailored output is markdown. pipe to pandoc for pdf, or paste into overleaf if you have a tex template.
• the model won't invent skills. if the base resume is thin, the tailored version is thin. fix the base first.

roadmap

• a built-in pdf renderer (drop the pandoc step)
• finch apply to autofill the application after tailoring, same engine as the web app
• optional sign-in to view your real applyfinch.com pipeline from the terminal

license

proprietary. personal + evaluation use only. see LICENSE for the full terms.