go-apply

AI-powered job application CLI. Scores your resume against job postings, tailors it, and generates cover letters.

Modes

Mode	Command	Use case
MCP Server	go-apply serve	Claude Code, Hermes, Openclaw — full orchestrated experience with automatic profile use
Headless CLI	go-apply run --url <url>	For agents without MCP support

MCP is the recommended path. The host agent drives the full tool flow and your onboarded profile is used automatically. For agents without MCP, use go-apply run directly — see Commands.

Quick Start

1. Install

   curl -sSfL https://raw.githubusercontent.com/thedandano/go-apply/main/scripts/install.sh | bash

2. Register with your MCP agent

   go-apply setup mcp --agent claude    # Claude Code
   go-apply setup mcp --agent hermes    # Hermes
   go-apply setup mcp --agent openclaw  # Openclaw

3. Ask your agent to onboard and apply — the agent drives the full flow via MCP tools:

   • "Onboard my resume at ~/docs/resume.md"
   • "Score my resume against this job posting and tailor it"

Installation

Install script (recommended)

curl -sSfL https://raw.githubusercontent.com/thedandano/go-apply/main/scripts/install.sh | bash

Install options (env vars):

# Install a specific version
curl -sSfL https://raw.githubusercontent.com/thedandano/go-apply/main/scripts/install.sh | VERSION=0.1.0 bash

# Install to a custom directory (e.g., system-wide, requires sudo)
curl -sSfL https://raw.githubusercontent.com/thedandano/go-apply/main/scripts/install.sh | INSTALL_DIR=/usr/local/bin sudo bash

To update, re-run the install command — it overwrites the existing binary.

Uninstall

# Remove binary only (keeps config, data, and logs)
curl -sSfL https://raw.githubusercontent.com/thedandano/go-apply/main/scripts/install.sh | bash -s -- --uninstall

# Remove everything (binary + config + data + logs)
curl -sSfL https://raw.githubusercontent.com/thedandano/go-apply/main/scripts/install.sh | bash -s -- --uninstall --purge

Configuration

Config file: ~/.config/go-apply/config.yaml

# CLI mode only — not needed for MCP (the host agent is the orchestrator)
orchestrator:
  base_url: https://api.anthropic.com/v1
  model: claude-sonnet-4-6
  api_key: sk-ant-...        # or set GO_APPLY_API_KEY env var

years_of_experience: 7
default_seniority: senior
user_name: "Your Name"
occupation: "Software Engineer"
location: "San Francisco, CA"

All tunable scoring constants (weights, thresholds, limits) live in internal/config/defaults.json.

Logging

Log files are written to ~/.local/state/go-apply/logs/go-apply-YYYY-MM-DD.log — one file per calendar day (multiple invocations append); last 50 files retained.

# View recent logs
go-apply logs

# Watch live (tail -f equivalent)
go-apply logs --follow

# Show last 200 lines
go-apply logs --lines 200

# Tail the raw log file with grep
tail -f ~/.local/state/go-apply/logs/go-apply-$(date +%Y-%m-%d).log | grep ERROR

Configuration

Log level and verbose mode are set in ~/.config/go-apply/config.yaml:

log_level: debug   # debug | info | warn | error (default: info)
verbose: true      # true = full request/response payloads in logs; false = truncated at 2 KB

Set them without editing the file directly:

go-apply config set log_level debug
go-apply config set verbose true

MCP server debug logging

To enable debug logs when running as an MCP server, set log_level in the config file, or pass it via the env block in Claude Code's settings.json (config file is read on startup):

{
  "mcpServers": {
    "go-apply": {
      "command": "go-apply",
      "args": ["serve"]
    }
  }
}

Run go-apply config set log_level debug once and every invocation — CLI and MCP — picks it up.

Commands

go-apply run

Run the full pipeline against a job description. Fetches (or accepts) the JD, scores all resumes in ~/.local/share/go-apply/inputs/, tailors resumes with a two-tier cascade (T1 keyword injection + T2 bullet rewriting), and generates a cover letter.

# From a URL (fetches and caches the JD)
go-apply run --url https://example.com/jobs/123

# From raw text (useful in scripts or when the page is paywalled)
go-apply run --text "We are hiring a senior Go engineer..."

# Specify application channel
go-apply run --url <url> --channel REFERRAL   # COLD (default), REFERRAL, RECRUITER

Output (stdout, JSON):

{
  "status": "success",
  "jd": { "title": "Senior Go Engineer", "company": "Acme", "required": ["go", "kubernetes"], ... },
  "scores": { "my-resume": { "breakdown": { "keyword_match": 40.5, ... }, ... } },
  "best_score": 82.3,
  "best_resume": "my-resume",
  "keywords": { "required": ["go", "kubernetes"], "preferred": ["docker"] },
  "cover_letter": { "text": "...", "channel": "COLD", "word_count": 180 },
  "start_time": "...", "end_time": "..."
}

Pipeline events (step-started/completed/failed) are written as JSON lines to stderr.

Flags:

Flag	Default	Description
--url	—	URL of the job posting
--text	—	Raw JD text (mutually exclusive with --url)
--headless	true	JSON output mode
--channel	COLD	Application channel: COLD, REFERRAL, RECRUITER
--accomplishments	—	Path to accomplishments doc; required for tailoring — omitting it skips T1 and T2 entirely

MCP Server (Claude Code, Hermes, Openclaw)

Add to Claude Code settings.json:

{
  "mcpServers": {
    "go-apply": { "command": "go-apply", "args": ["serve"] }
  }
}

Available tools: onboard_user, add_resume, get_config, update_config, load_jd, submit_keywords, submit_tailor_t1, submit_tailor_t2, finalize

MCP setup command

Instead of editing config files manually, use the setup command:

# Register with Claude Code
go-apply setup mcp --agent claude

# Register with OpenClaw
go-apply setup mcp --agent openclaw

# Register with Hermes
go-apply setup mcp --agent hermes

# Register with all known agents at once
go-apply setup mcp --agent all

To unregister:

go-apply setup mcp --agent claude --remove
go-apply setup mcp --agent all --remove

To overwrite an existing registration:

go-apply setup mcp --agent claude --override   # or --force (alias)

The command is idempotent — running it again without --override reports "already registered" and makes no changes. On a TTY you will be prompted to confirm the overwrite.

CLI Reference

Global (persistent) flags

These apply to every subcommand and must come before the subcommand name.

There are no persistent global flags. Log level and verbose mode are configured via go-apply config set (see Logging).

go-apply run

Run the full apply pipeline against a job description.

Flag	Default	Description
--url <url>	—	URL of the job posting to fetch
--text <jd>	—	Raw job description text (mutually exclusive with --url)
--headless	true	JSON output mode
--channel <channel>	COLD	Application channel: COLD, REFERRAL, RECRUITER
--accomplishments <path>	—	Path to accomplishments doc; required for tailoring — omitting it skips T1 and T2 entirely

go-apply serve

Start the MCP stdio server for Claude Code integration. No flags.

go-apply onboard

Store resumes, skills, and accomplishments in the profile database.

Flag	Default	Description
--resume <path>	—	Path to a resume file (repeatable; at least one required)
--skills <path>	—	Path to skills reference file (optional)
--accomplishments <path>	—	Path to accomplishments file (optional)
--reset	false	Delete profile database and inputs/ directory
--yes	false	Skip confirmation prompt for --reset (required for non-interactive use)

go-apply config

Manage go-apply configuration. Subcommands:

Subcommand	Usage	Description
set	go-apply config set <key> <value>	Set a config field by dot-notation key
get	go-apply config get <key>	Get a config field value by dot-notation key
show	go-apply config show	Show all config fields (API keys redacted)

Config keys use dot notation (e.g. llm.base_url, embedder.model, user_name, log_level).

go-apply setup mcp

Register or unregister go-apply as an MCP server in an AI agent's config.

Flag	Default	Description
--agent <name>	—	Agent to configure: claude, openclaw, hermes, all (required)
--remove	false	Unregister go-apply from the agent's config
--override	false	Overwrite an existing registration
--force	false	Alias for --override

go-apply logs

View recent go-apply log entries.

Flag	Shorthand	Default	Description
--lines <n>	-n	100	Number of recent lines to show
--follow	-f	false	Watch for new log lines (tail -f mode)

go-apply update

Update go-apply to the latest GitHub release. No flags.

go-apply version

Print the go-apply version. No flags.

Roadmap

Feature	Status
Resume tailoring (keyword injection + bullet rewriting)	Shipped
Multi-resume scoring with full breakdown	Shipped
MCP integration (Claude Code, Hermes, Openclaw)	Shipped
Headless CLI (agents without MCP)	Shipped