Helmdeck
Built for teams running local or airgapped models who need browser automation, desktop control, and code workflows without sending credentials to cloud APIs. Ships 52 capability packs as typed JSON tools that 7B–30B models can call reliably: browser sessions, git operations, vision tasks, and multi-step pipelines all wrapped behind single MCP calls. The pitch is cost and control: a six-step code edit loop runs for $0.07 on gpt-oss-120b versus $0.30+ on Sonnet, and everything stays in your VPC. Includes routing memory, intent decomposition, and a management UI. If you're optimizing for small model success rates or can't let an LLM see your SaaS tokens, this is the stack.
- GitHub stars
- 6
- tools
- 37
- auth
- Secret needed
- transport
- STDIO
- env vars
- 2
- last updated
- Jun 22, 2026
Tools
37 toolsPublic tool metadata for what this MCP can expose to an agent.
browser.screenshot_urlSession lifecycle, navigation, render wait, cleanup
Session lifecycle, navigation, render wait, cleanup
No parameter schema in public metadata yet.
browser.interactDeterministic multi-step CDP (navigate, click, type, scroll, screenshot, assert_text) — no LLM needed
Deterministic multi-step CDP (navigate, click, type, scroll, screenshot, assert_text) — no LLM needed
No parameter schema in public metadata yet.
web.scrapeFirecrawl-backed markdown scrape
Firecrawl-backed markdown scrape
No parameter schema in public metadata yet.
web.scrape_spaSchema-driven SPA extraction
Schema-driven SPA extraction
No parameter schema in public metadata yet.
web.testNatural-language browser tests via Playwright MCP + LLM loop
Natural-language browser tests via Playwright MCP + LLM loop
No parameter schema in public metadata yet.
research.deepMulti-source Firecrawl search + per-source scrape + LLM synthesis with inline citations
Multi-source Firecrawl search + per-source scrape + LLM synthesis with inline citations
No parameter schema in public metadata yet.
content.groundParses a markdown file for claims, finds authoritative sources, inserts real citations in place
Parses a markdown file for claims, finds authoritative sources, inserts real citations in place
No parameter schema in public metadata yet.
slides.renderMarp + Chromium + format flags
Marp + Chromium + format flags
No parameter schema in public metadata yet.
slides.narrateNarrated MP4 video with ElevenLabs TTS per slide, YouTube metadata, sidecar SRT captions and structured validation
Narrated MP4 video with ElevenLabs TTS per slide, YouTube metadata, sidecar SRT captions and structured validation
No parameter schema in public metadata yet.
av.validateStructured AV-artifact validation of faststart, codec, packet contiguity, RMS, LUFS, duration and SRT format
Structured AV-artifact validation of faststart, codec, packet contiguity, RMS, LUFS, duration and SRT format
No parameter schema in public metadata yet.
doc.parseDocling layout-aware parse of PDF tables, multi-format, OCR fallback
Docling layout-aware parse of PDF tables, multi-format, OCR fallback
No parameter schema in public metadata yet.
doc.ocrTesseract fallback for simple images
Tesseract fallback for simple images
No parameter schema in public metadata yet.
desktop.run_app_and_screenshotXvfb + xdotool + scrot + window focus
Xvfb + xdotool + scrot + window focus
No parameter schema in public metadata yet.
vision.click_anywhereNative computer-use routing with JSON-prompt fallback for Ollama/Deepseek
Native computer-use routing with JSON-prompt fallback for Ollama/Deepseek
No parameter schema in public metadata yet.
vision.extract_visible_textScreenshot to vision model to action loop
Screenshot to vision model to action loop
No parameter schema in public metadata yet.
vision.fill_form_by_labelScreenshot to vision model to action loop
Screenshot to vision model to action loop
No parameter schema in public metadata yet.
repo.fetchSSH key selection from vault, known_hosts, key shred-on-exit; returns tree, readme, entrypoints and signals
SSH key selection from vault, known_hosts, key shred-on-exit; returns tree, readme, entrypoints and signals
No parameter schema in public metadata yet.
repo.pushSSH key selection from vault, known_hosts, key shred-on-exit
SSH key selection from vault, known_hosts, key shred-on-exit
No parameter schema in public metadata yet.
repo.mapAider-style structural symbol map under a token budget
Aider-style structural symbol map under a token budget
No parameter schema in public metadata yet.
fs.readPath-safe file operations inside a clone
Path-safe file operations inside a clone
No parameter schema in public metadata yet.
fs.writePath-safe file operations inside a clone
Path-safe file operations inside a clone
No parameter schema in public metadata yet.
fs.patchPath-safe file operations inside a clone
Path-safe file operations inside a clone
No parameter schema in public metadata yet.
fs.listPath-safe file operations inside a clone
Path-safe file operations inside a clone
No parameter schema in public metadata yet.
fs.deletePath-safe file operations inside a clone
Path-safe file operations inside a clone
No parameter schema in public metadata yet.
cmd.runRun an arbitrary command in a clone path
Run an arbitrary command in a clone path
No parameter schema in public metadata yet.
git.commitStage, commit and review changes attributed to helmdeck-agent
Stage, commit and review changes attributed to helmdeck-agent
No parameter schema in public metadata yet.
git.diffStage, commit and review changes attributed to helmdeck-agent
Stage, commit and review changes attributed to helmdeck-agent
No parameter schema in public metadata yet.
git.logStage, commit and review changes attributed to helmdeck-agent
Stage, commit and review changes attributed to helmdeck-agent
No parameter schema in public metadata yet.
github.create_issueVault-stored PAT, never visible to the agent
Vault-stored PAT, never visible to the agent
No parameter schema in public metadata yet.
github.list_issuesVault-stored PAT, never visible to the agent
Vault-stored PAT, never visible to the agent
No parameter schema in public metadata yet.
github.list_prsVault-stored PAT, never visible to the agent
Vault-stored PAT, never visible to the agent
No parameter schema in public metadata yet.
github.post_commentVault-stored PAT, never visible to the agent
Vault-stored PAT, never visible to the agent
No parameter schema in public metadata yet.
github.create_releaseVault-stored PAT, never visible to the agent
Vault-stored PAT, never visible to the agent
No parameter schema in public metadata yet.
github.searchVault-stored PAT, never visible to the agent
Vault-stored PAT, never visible to the agent
No parameter schema in public metadata yet.
python.runCPython 3 + pytest + ruff + mypy in a Python sidecar image
CPython 3 + pytest + ruff + mypy in a Python sidecar image
No parameter schema in public metadata yet.
node.runNode 20 LTS + npm + pnpm + yarn + tsc in a Node sidecar image
Node 20 LTS + npm + pnpm + yarn + tsc in a Node sidecar image
No parameter schema in public metadata yet.
http.fetchPlaceholder-token egress with vault token substitution in URL, headers and body
Placeholder-token egress with vault token substitution in URL, headers and body
No parameter schema in public metadata yet.
Sponsors
Agent CaffeineSponsoredClose your lid, come back to finished work
SnugSponsoredThe clipboard manager for the AI eraPromote your product hereReach 125k+ visitors monthly
WaxSponsoredBlock anything with supervision, without erasing your iPhonePromote your product hereReach 125k+ visitors monthly
knobsSponsoredSee your site on any phone, any day, in any languageConfiguration
HELMDECK_URL*Base URL of your helmdeck control plane (e.g. http://localhost:3000 for a local Compose install, or your reverse-proxied https URL)
HELMDECK_TOKEN*secretJWT bearer token minted from your helmdeck control plane. Generate with: curl -X POST $HELMDECK_URL/api/v1/auth/login -d '{"username":"admin","password":"..."}'
Readme
helmdeck
Today's helmdeck install ran a full 6-step code-edit loop (clone, read, patch, test, commit, push) on
gpt-oss-120bfor $0.07. The same loop on Cursor or Claude Code direct via Sonnet would have cost $0.30+. Same outcome, ~5× cheaper — and the "expensive" stack isn't even the most expensive option.
| Workflow | Frontier-model approach | Helmdeck (gpt-oss-120b) |
|---|---|---|
| Browser scrape + GitHub comment | $0.25 (Anthropic Computer Use) | $0.005 |
| Code edit loop (6 steps) | $0.35 (Cursor / Aider) | $0.07 |
| Multi-step browser test | $0.20 (Browser-use NL) | $0.03 |
| PDF → structured Markdown | $1.00 (naive Sonnet vision) | $0.003 |
Details
- Repository
- github.com/tosin2013/helmdeck
- Website
- helmdeck.dev
- Categories
- Web & Browser Automation, AI & LLM Tools
- Package
@helmdeck/mcp-bridge- Transport
- STDIO
- Auth
- Required
- Registry
- MCP registry, active
More from tosin2013
Related Web & Browser Automation MCP servers
View all
Argusironclawdevs27/argusAutomated web QA via Chrome DevTools MCP — a11y, security, visual, env diff, CI; Aegis redaction.5stars
Pagerunner — The Browser Layer That Learnsenreign/pagerunnerReal Chrome profiles with site intelligence — already logged in, PII anonymization, sessions persist3stars
Anybrowsekc23go/anybrowseConverts any URL to clean, LLM-ready Markdown using real Chrome browsers3stars
anysitedwytmcl/anysite-mcpAny website as MCP tools: a real Chrome signed in as you, and shortcuts the agent saves as tools.1stars
Powershell Mcpimrrd/powershell-mcpHeadless Windows ops over MCP: hidden PowerShell plus in-process SSH, WinRM and SFTP.1stars
Assiststudio Runnerfieldcure/fieldcure-assiststudio-runnerWindows-only headless LLM task automation engine with scheduling via Windows Task Scheduler.1stars