使用说明

go_job

Job search, resume optimization, career research, and application tracking — a Go MCP server exposing 28 tools across LinkedIn, Greenhouse, Lever, YC, HN, Indeed, Хабр, RemoteOK, WeWorkRemotely, Twitter/X, Google Jobs, and more.

MCP Tools

Search

Tool	Description
`job_search`	Unified search: LinkedIn, Greenhouse, Lever, YC, HN, Indeed, Хабр, RemoteOK, WeWorkRemotely, Remotive, Twitter/X, Google Jobs. `platform=` selects source; `limit` (default 15, max 50) + `offset` for pagination.
`job_match_score`	Score job listings against a resume using Jaccard keyword overlap (0–100).
`opportunity_search`	Cross-type opportunity search (jobs + freelance + bounty).
`opportunity_analyze`	Deep analyze a single opportunity URL.
`opportunity_claim`	Initiate a claim action on a matched opportunity.

Resume

Tool	Description
`resume_analyze`	ATS score (0–100), missing keywords, gaps, recommendations.
`cover_letter_generate`	Tailored cover letter (3 tones: professional / friendly / concise).
`resume_tailor`	Rewrite resume sections to match JD, keyword diff.
`master_resume_build`	Build a master resume profile from raw experience.
`resume_generate`	Generate a targeted resume from the master profile.
`resume_enrich`	Enrich master profile via Q&A.
`resume_profile`	View the stored master profile.
`resume_memory`	Semantic search/add/update over resume memory store.

Research

Tool	Description
`research`	Research salary, company, or person. `subject=salary\|company\|person`.
`ats`	Direct ATS board fetch (Greenhouse, Lever, Ashby).

Interview & Career Prep

Tool	Description
`interview_prep`	Personalized Q&A (behavioral + technical + system design) with model answers.
`project_showcase`	STAR-format project narratives with impact and talking points.
`pitch_generate`	30-sec & 2-min elevator pitches, "why this company" answer.
`skill_gap`	Resume vs JD gap: match score, missing skills, learning plan.

Application Workflow

Tool	Description
`application_prep`	One-call combo: resume analysis + cover letter + interview prep + company research.
`offer_compare`	Side-by-side offer comparison with scoring (0–100).
`negotiation_prep`	Salary negotiation playbook: scripts, counters, BATNA.
`linkedin`	LinkedIn profile operations.
`linkedin_profile_ingest`	Ingest a LinkedIn profile for local analysis.

Tracker & Utilities

Tool	Description
`job_tracker`	Track job applications. `action=add\|list\|update`.
`algora_job_ingest`	Ingest Algora bounty/job listings into the hunt store.
`hunt_list`	List hunt entries from the local store (triggers lazy enrichment).
`oversize`	Retrieve / list / purge oversized MCP responses from the spillover store.

Filters (job_search)

Filter	Values
`experience`	internship, entry, associate, mid-senior, director, executive
`job_type`	full-time, part-time, contract, temporary
`remote`	onsite, hybrid, remote
`time_range`	day, week, month
`salary`	40k+, 60k+, 80k+, 100k+, 120k+, 140k+, 160k+, 180k+, 200k+
`easy_apply`	true (LinkedIn Easy Apply only)
`platform`	linkedin, greenhouse, lever, ats, yc, hn, indeed, habr, remoteok, weworkremotely, remotive, twitter, google, un, inspira, undp, all (default)
`limit`	1–50 (default 15)
`offset`	skip N results for pagination
`blacklist`	comma-separated company/keyword exclusion

Architecture

go_job/
├── main.go
├── internal/
│   ├── engine/
│   │   ├── config.go          # Config struct + Init()
│   │   ├── bridge.go          # Source bridge wiring
│   │   ├── bridge_jobs.go     # Job-specific bridge helpers
│   │   ├── bridge_llm.go      # LLM bridge
│   │   ├── cache.go           # 2-tier cache: L1 in-memory + L2 Redis
│   │   ├── search.go          # SearchSearXNG, DedupByDomain
│   │   ├── metrics.go         # Prometheus counters/histograms
│   │   ├── pipeline.go        # Fan-out pipeline helpers
│   │   ├── types_jobs.go      # Input/output types
│   │   ├── prompt.go          # LLM instructions (shared)
│   │   ├── prompt_jobs.go     # Job-specific LLM prompts
│   │   └── jobs/              # Job source + career tool implementations
│   │       ├── linkedin.go    # LinkedIn Guest API + JSON-LD + geo_id + pagination
│   │       ├── indeed.go      # Indeed iOS GraphQL API + SearXNG fallback
│   │       ├── remotejobs.go  # RemoteOK + WeWorkRemotely + Remotive APIs
│   │       ├── habr.go        # Habr Карьера scraper
│   │       ├── hnjobs.go      # HN Who is Hiring (Algolia)
│   │       ├── ycjobs.go      # YC workatastartup.com
│   │       ├── ats.go         # Greenhouse + Lever + Ashby ATSes
│   │       ├── match.go       # Jaccard keyword scoring (job_match_score)
│   │       ├── resume.go      # resume_analyze, cover_letter_generate, resume_tailor
│   │       ├── research.go    # research tool (salary / company / person)
│   │       ├── tracker.go     # Job application tracker (SQLite)
│   │       ├── profile.go     # User profile persistence
│   │       └── ...            # algora, twitter, linkedin, bounty, opportunity, etc.
│   │   └── sources/           # Pluggable source connectors
│   │       ├── freelancer.go  # Freelancer.com REST API
│   │       ├── github.go      # GitHub jobs / PRs
│   │       ├── hackernews.go  # HN source
│   │       └── ...
│   ├── jobserver/
│   │   ├── register.go        # Tool registrations (28 MCP tools)
│   │   └── tool_*.go          # Per-tool handler files
│   └── hunt/                  # Hunt store + notifications
│       ├── notify/
│       │   └── telegram.go    # Telegram notifications via go-kit ProductSink
│       └── ...
└── deploy/
    └── go_job.service         # systemd unit (MCP :8891, metrics :9891)

Key Implementation Details

job_search

Unified platform: remote_work_search and freelance_search and twitter_job_search are folded into job_search via platform=remoteok|weworkremotely|remotive|twitter. Use raw=true with platform=twitter to skip LLM processing.
LinkedIn: no auth, Chrome TLS fingerprint via bogdanfinn/tls-client; pagination with offset; 42 geo locations.
Google Jobs: platform=google via SearXNG.
UN sources: platform=inspira (careers.un.org) / platform=undp / platform=un (fan-out both). Not included in platform=all.

research tool

Replaces three separate tools (salary_research, company_research, person_research).
subject=salary (role required), subject=company (company required), subject=person (name required).

job_tracker

Single tool replaces job_tracker_add, job_tracker_list, job_tracker_update.
action=add (title+company required), action=list, action=update (id required).

Data Storage

Job tracker DB: $UPLOADS_ROOT/go-job/tracker/tracker.db (SQLite, table jobs). Default path: $HOME/uploads/go-job/tracker/tracker.db. Override via UPLOADS_ROOT.
User profile: $UPLOADS_ROOT/go-job/profile/profile.json. Default: $HOME/uploads/go-job/profile/profile.json.
L1 cache: in-memory (sync.Map), lost on restart.
L2 cache: Redis (optional), persistent.

Notifications

New hunt entries trigger Telegram notifications via internal/hunt/notify/telegram.go using go-kit ProductSink (own bot, rate-limited fan-out).
Requires TELEGRAM_BOT_TOKEN + HUNT_NOTIFY_CHAT_ID env vars.

Running

# HTTP mode (default MCP port 8891, metrics port 9891)
MCP_PORT=8891 PROM_PORT=9891 LLM_API_KEY=... ./bin/go_job

# stdio mode
./bin/go_job --stdio

Build & Deploy

make build    # → bin/go_job
make deploy   # build + copy service + restart systemd unit
make restart  # restart only

Config (env vars)

Var	Default	Description
`SEARXNG_URL`	`http://127.0.0.1:8888`	SearXNG instance
`LLM_API_KEY`	(required)	API key for the OpenAI-compatible LLM gateway
`LLM_API_BASE`	`http://127.0.0.1:8317/v1`	OpenAI-compatible LLM gateway (cliproxyapi) base URL
`LLM_MODEL`	(deploy-set)	Primary model name. Model selection is dynamic (go-engine/llm) — see `LLM_MODEL_FALLBACK`
`LLM_MODEL_FALLBACK`	(empty)	CSV cross-provider fallback chain: tries the primary, then each entry on retryable failure, health-filtered against the gateway's `/v1/models`
`MCP_PORT`	`8891`	MCP HTTP server port
`PROM_PORT`	`9891`	Prometheus metrics port
`REDIS_URL`	(optional)	Redis for L2 cache
`CACHE_TTL`	`900`	Cache TTL in seconds
`UPLOADS_ROOT`	`$HOME/uploads`	Base dir for tracker DB + user profile
`DATABASE_URL`	(optional)	Postgres for oversized payload spillover
`TELEGRAM_BOT_TOKEN`	(required for notifications)	Telegram bot token
`HUNT_NOTIFY_CHAT_ID`	(required for notifications)	Notification recipient chat ID

Health check

curl http://localhost:8891/health
# {"status":"ok","service":"go_job","version":"v1.21.0"}
#
# `version` is `git describe --tags` taken at build time — the release tag that
# release-please cut, never a number maintained by hand here. A build that
# cannot reach git reports "dev".

Metrics

curl http://localhost:9891/metrics

Spillover store

Large MCP responses can exceed the MCP envelope limit (~25KB). go-job spills oversized payloads to a Postgres table oversize_responses (auto-migrated on startup).

When a tool response exceeds GO_JOB_OVERSIZE_THRESHOLD_BYTES (default 24576), the client receives a small envelope with oversize_id and a sample. Use the oversize tool to retrieve the full payload.

Requires DATABASE_URL. Gracefully falls back to direct response if unset.