In July, my last company let over 200 people go due to a shift in direction. Unfortunately, I was part of that and that same day I created TruthCV to share with my colleagues to help on their job hunt.
I developed it agentically with Forestree which puts architecture first.
How I use it
After set up, I let it find applications and I just review them with my morning coffee. I've mainly been using it as a filter to only find the jobs I'd be really interested in reviewing. If it brings me any I go to the site and apply manually.
It has a feature to generate cover letters which I use to spit out a base doc then manually edit as I see fit.
Finally, it helps with keeping track of the applications you've put in. I usually add little notes about what salary range I requested, whether I reached out to some one personally, etc.
High Level Instructions
You load your CV, set up your job search and schedule and just let it run on a server or whatever machine you have running 24/7.
TruthCV searches five job boards out of the box (Greenhouse, Lever, Ashby, Workday and Arbeitnow). You can add any other site on the Job boards page. Give it the domain, and optionally a search link with {keywords} and {location} slots. The agent then either searches the site directly or finds its postings through Google.
Some sites need a login. The agent uses its own built-in browser, so you sign in once through that browser inside the app. TruthCV remembers the login. If the agent ever hits a login wall, the site shows up under "Needs attention" with a one-click "Sign in" button.
For deploying TruthCV
Harness users: You can use your own agents to set this up by telling them to check the README for instructions.
Manual set up can be done by following the README
Must haves
- Docker, installed and running. On Mac or Windows that means Docker Desktop; on Linux, Docker plus Compose.
- One AI model connection:
- Claude: a Pro/Max subscription sign-in, or an Anthropic API key Didn't test the following:
- ChatGPT: a subscription sign-in, or an OpenAI API key
- OpenRouter: an API key
- Ollama: free and local, no key needed Your CV as a PDF, DOCX, TXT or Markdown file. A LinkedIn PDF export is the easiest.
Good to know
- Install: double-click the launcher for your system in scripts/launch/ (.command for Mac, .bat for Windows, .sh for Linux). No config file to write.
- First start takes about 10 minutes, because it builds the app once. After that it starts in seconds.
- It runs 3 containers: the app, a Chromium browser and the job-application agent. The browser container reserves 2 GB of shared memory. Ollama is a 4th container, only if they choose it.
- Data stays on your machine in ./data. There's no database and no account to create.
- Optional: Gmail response tracking (needs your own Google Cloud OAuth app, plus a Jev key) and Jev cross-checking (needs a Jev key).
The app has no login. It only listens on the machine itself (port 5627), and that's on purpose. If you put it on a server or the open internet, you must add your own login in front of it. I use mine over vpn when I'm away from home.
Hope it helps with the daily grind, I've gotten a few interviews by using it but nothing has landed yet ¯\(ツ)/¯
Top comments (0)