meet opengrasp

Open-source terminal agent - discovers roles, scores fit, tailors ATS-safe CVs, and drafts applications entirely on your machine.

$pip install opengrasp

Read the Docs

opengrasp - zsh

Live

manas@local:~/opengrasp$ █

step0/4

inferenceollama

statusready

How it works

End-to-End Local Pipeline

From job discovery to application - entirely on your machine. Nothing leaves.

opengrasp pipeline --list-steps

6 stages

$ opengrasp pipeline --describe

[01]Job DiscoveryScans portals, deduplicates against DB & history

[02]10-Dimension Fit ScoringRole, skills, seniority, comp, geo, stage, PMF, growth, interview, timeline

[03]ATS CV GenerationKeyword extraction, bullet reordering, PDF via Playwright

[04]Human-Reviewed DraftingPrefills form fields — never auto-submits

[05]Parallel Batch ProcessingConcurrent queue, fault tolerant, resumable

[06]Outcome Learning LoopLogs results, adjusts scoring weights over time

$ opengrasp scan --auto # Full automated scan

$ opengrasp batch --min-score B --limit 20 # Batch

HITL AlwaysPrivacy-first

● All data stays local · SQLite DB · No cloud uploads

Why opengrasp

Built for serious job seekers

opengrasp makes it accessible to everyone.

Privacy-First

CV, config, reports live locally. SQLite DB on your machine. Prompt execution via local Ollama — zero data sent to cloud.

$opengrasp setup

Fit Scoring

Scores each role 1.0–5.0 across 10 dimensions. Grades A/B/C/D/F. Writes a markdown report to reports/.

$opengrasp scan

ATS CV Generation

Extracts JD keywords, reorders bullets by relevance without deleting your content. Renders ATS-safe HTML to PDF.

$opengrasp pipeline <url>

HITL Philosophy

AI analyzes. You decide. The system drafts and prefills — but a human must review before any application is marked applied.

$opengrasp apply <url>

Batch Processing

Process queued URLs concurrently. Fault tolerant — one failure doesn't block the queue. Fully resumable.

$opengrasp batch --min-score B

Outcome Learning

Log interview/rejection/offer/ghosted outcomes. Scoring weights in DB auto-adjust based on feedback patterns.

$opengrasp learn <id> offer

Inference

Local Ollama Models

All inference runs locally through Ollama. No API keys. No subscriptions. No tracking.

opengrasp model --list

Ollama local

$ ollama list

[1]●llama3.1:8b(evaluation)

Fast, low footprint — scoring & routing

[2]●qwen2.5:14b(generation)

Stronger CV/cover-letter drafting quality

[3]●any-local-model(custom)

Configurable via portals.yml

Guidance:

$ ollama pull llama3.1:8b # scan/eval

$ ollama pull qwen2.5:14b # CV/letter gen

$ opengrasp doctor # verify setup

100% LocalZero cloud

● Use smaller models for volume loops · Larger models for final artifact quality

Get started

Ready to own your job search?

Install in two commands. Run entirely locally. Your data, your machine, your control.

Quick start

$pip install opengrasp

$opengrasp setup

$opengrasp doctor

$opengrasp scan --limit 5 --link-limit 30

Star on GitHub

$ pip install opengrasp