- Home
- Skills
- Workflow Automation
- anybridge
Works with the AI Agents you already use
anybridge
Automate API-less software using an MCP server for deterministic, human-approved, and verified workflows.
Free
anybridge
Example session with this skill installed
I need to add a new task called 'Review PR' to my local task manager at http://localhost:4010 and make sure it actually saved to the data file.
- Read your context and instructions
- Compiled the anybridge
I have inspected the task manager interface and proposed a workflow to add the task.
Proposed Workflow: Add Task
| Step | Action ID | Description | Expected State |
|---|---|---|---|
| 1 | a1 | Open task input field | opened |
| 2 | a8 | Input value "Review PR" | filled |
| 3 | a9 | Click "Add Task" button | added |
Verification Check
- Kind:
file_contains - Target:
./data/tasks.json - Expected Value: "Review PR"
Risk Summary
This workflow performs a write operation to a local web origin and modifies a local JSON file.
Next steps
- Approve this plan in your terminal by running:
node dist/index.js approve plan_8b2f92d1c5e4 - Once approved, tell me to execute the workflow.
- After execution, I will verify the file content and we can save this as a reusable workflow.
Connects securely to your tools. The creator never sees your data.
About this skill
Automating software without an API usually forces developers into a cycle of fragile scripts, brittle CSS selectors, and high model token costs. AnyBridge solves this by providing a local MCP server that transforms CLI tools and web interfaces into a deterministic, numbered menu system.
What it does
- Interface discovery identifies allowlisted CLI tools, local web origins, and saved workflows.
- Menu-based inspection generates a numbered list of valid action IDs, preventing the agent from inventing selectors or commands.
- Workflow compilation allows the agent to propose a multi-step execution plan including mandatory verification checks.
- Deterministic replay executes saved workflows with zero model calls using hash-pinned verification.
- Out-of-band approval enforces a human-in-the-loop requirement where the user must manually approve plans in the terminal.
How it works
- Discover & Inspect the target software to receive a menu of valid IDs and their side effects.
- Propose a plan containing a sequence of actions and at least one independent verification check.
- Approve the plan via a local terminal command using the provided
plan_idand hash. - Execute & Verify the plan, where the system runs the steps and confirms the outcome via file or DOM inspection.
- Save & Replay successful plans as parameterized workflows for future use without LLM intervention.
Frameworks & tools
This skill requires Node.js 20+, npm, and Chromium via Playwright. It functions as a local Model Context Protocol (MCP) server compatible with Claude Code, Cursor, and VS Code.
Why this beats prompting it yourself
Raw prompting leads to hallucinated CLI flags and broken web selectors that fail silently. This skill enforces a strict contract of numbered IDs, mandatory verification checks, and human approval, ensuring operations are both safe and repeatable without wasting tokens on every run.
Use cases
- Automating internal admin panels or legacy web apps lacking a public API.
- Executing repeatable CLI tasks with proof of success via file hash or exit code.
- Converting one-off agent actions into reusable, verified automation scripts.
- Operating local development tools that require strict human oversight before execution.
Known limitations
Restricted to CLI and Chromium-based browser targets within a local allowlist. Desktop UI automation is not supported, and it does not serve as a security sandbox for untrusted targets.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Workflow Automation.
CLI and Chromium targets only. Requires local Node 20+ and manual terminal approval. Only operates on targets in the explicit allowlist. No desktop app support.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 1 day ago
- Free with an Agensi account
- Software
Node.jsPlaywright
Git