eric-task-master
Sends notifications via a Telegram Bot when a live task requires a human verification click, repeating every 30 seconds until the request is claimed.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@eric-task-masterLaunch an unattended job to log into the portal and download reports"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Eric Task Master 3.00
One task. One small script. One command.
Eric Task Master is an always-on local Chrome task system for AI agents. It gives Codex, Claude Code, WorkBuddy, Hermes, Pi, and any other local Agent with terminal access a shared way to start long-running browser work, stream partial results, and leave the task running after the Agent disconnects.
Why
Built-in Agent browsers are short-lived. Direct browser control is flexible but usually loses task state, progress, and cleanup when an Agent turn ends. Task Master keeps that operational layer running while leaving the actual browser logic completely free.
Fast start: write the smallest useful
.mjs, then run one command.Full browser freedom: use Playwright,
page.evaluate(), CDP, HTTP, files, and your own retry logic directly.Durable work: the Manager and Worker continue after the calling Agent exits.
Real signed-in sessions: named persistent Profiles use the computer's stable Chrome.
Multi-Agent ready: different Profiles run concurrently; one Profile has one writer at a time.
Results while running: progress and output files remain useful even when a task stops early.
Verification without losing the task: keep Chrome open, request a system notification immediately and every 30 seconds until actual resume, and take screenshots at 5/10/15/20 minutes. After 20 minutes, reminders stop and the task stays automatically paused until manual resume.
Simple recovery: stale leases are reclaimed; stop and delete terminate owned processes and release the Profile.
Quiet local operation: CLI only, loopback only. No MCP registration, browser extension, pairing code, or task-asset catalog.
Use it for unattended research, recurring operations, long browser collections, account workflows, form work, monitoring, content operations, QA, or any custom web process an Agent can express in JavaScript.
Related MCP server: OpenChrome
Install
Download two files from the latest GitHub Release:
the Manager installer or portable ZIP matching your OS and CPU;
eric-task-master-skill-v3.1.2.zip.
The Manager package includes its Node.js runtime, Playwright runtime, CLI, Dashboard, and background service. It uses an installed stable Google Chrome and does not download a separate Chromium.
Installer unavailable? Download eric-task-master-v3.1.2-<target>-portable.zip, verify it against SHA256SUMS, and extract it to a permanent folder. Run eric-task-master/bin/taskmaster.cmd panel on Windows or eric-task-master/bin/taskmaster panel on macOS/Linux. No installer, administrator access, or separate Node.js installation is needed. Choose your target and view setup details.
After installation:
taskmaster panelCreate a Profile, sign in in its native Chrome window, close that window, and set it as default. Tasks reuse the same Profile. Then give the Skill to the Agent and state the task. Sites may still ask for verification again.
The Dashboard shows only current tasks, including queued, running, waiting, and stopping work. Ended tasks leave the list automatically; results and files remain available through the CLI.
The Dashboard's Clean space dialog clears idle Profile browser caches and finished-task temporary scripts. Historical screenshots, downloads, and results require a separate opt-in. Login state, extension data, and active tasks are preserved.
Agent path
taskmaster run ./job.mjs --input '@./input.json' --detach --json
taskmaster follow TASK_ID --jsonOmit --profile to use the default Profile. run starts the local Manager automatically when needed.
Minimal task:
export async function run({ page, input, outputDir, progress, signal }) {
await page.goto(input.url, { waitUntil: 'domcontentloaded' });
const title = await page.title();
await progress({ current: 1, total: 1, message: 'Collected title' });
return { title, outputDir, aborted: signal.aborted };
}The script receives the real Playwright page, context, and browser. It may use direct Playwright APIs, evaluate JavaScript, open CDP sessions, perform HTTP requests, and write incremental results to outputDir.
The submitted .mjs entry is the only file frozen into the task directory. Keep it self-contained: use Node built-ins, bare playwright, task input, absolute paths, or outputDir; relative sibling imports and resources beside the Agent's source file are not copied.
Essential commands
taskmaster status --json
taskmaster profiles --json
taskmaster run ./job.mjs --input '@./input.json' --json
taskmaster follow TASK_ID --json
taskmaster stop TASK_ID --json
taskmaster resume TASK_ID --json
taskmaster delete TASK_ID --json
taskmaster panelTrust model
Task scripts run as the current operating-system user and are trusted local code. Task Master does not pretend to sandbox them. Run scripts and Skills only from sources you trust.
Task Master itself binds to 127.0.0.1, redacts credentials from Manager-owned diagnostics, never automatically replays an entire failed script, and never launches two writers against the same Profile directory.
Project principle
Every design decision is judged against one path:
Receive the task → write the smallest free script → run one CLI command → stream useful results.
If a feature adds a required step before the first browser action without improving process safety or durability, it does not belong in the Manager.
License
MIT
Keywords: AI agent browser automation, Playwright automation, Chrome automation, long-running agent tasks, unattended automation, local AI agent, browser profile manager, CLI automation, multi-agent automation, RPA, web research, data collection.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections. Dates show when Glama detected each change.
No tool schema history has been recorded yet.
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Connectors
- openhelmOAuthai.openhelm
Autonomous cloud agent tasks: real browser + your tools, structured evidence-backed results.
Stealth web automation for AI agents. Login, signup, navigate, screenshot.
Stealth web automation for AI agents. Login, signup, navigate, screenshot.
Run multi-step tasks in a real Chrome browser: persistent environments, live view, human takeover.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceEnables AI assistants to control web browsers through Playwright automation, providing 50+ tools for navigation, interaction, testing, accessibility audits, and visual testing across Chromium, Firefox, and WebKit.15MIT
- AlicenseBqualityAmaintenanceControls a real Chrome browser for AI agents, enabling authenticated automation with parallel lanes, token-efficient page reads, and robust recovery mechanisms.100114236MIT
- AlicenseNot gradedqualityDmaintenanceEnables browser automation for AI assistants via Playwright, supporting multiple browsers, sessions, and tools for web interaction and testing.1,353MIT
- FlicenseNot gradedqualityCmaintenanceEnables AI assistants to perform web automation tasks such as navigation, typing, clicking, and taking screenshots using Playwright in Cloudflare Workers.-
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/npcworkspace-cmyk/eric-task-master'
If you have feedback or need assistance with the MCP directory API, please join our Discord server