Connect any MCP-enabled AI agent

lokeye exposes a remote MCP server. Connect Claude, Cursor, or any MCP client to dispatch real-world tasks, read live context, and trigger structured workflows — directly from your AI agent.

Setup
Configure your MCP client
Add lokeye as a remote MCP server. You need an API key — register here if you don't have one.

Cursor / VS Code

Add to your .cursor/mcp.json or VS Code MCP settings:

{
  "mcpServers": {
    "lokeye": {
      "type": "http",
      "url": "https://www.lokeye.in/mcp",
      "headers": {
        "X-API-Key": "YOUR_API_KEY"
      }
    }
  }
}

Claude Code (CLI)

Add the remote MCP server via the CLI. Works on macOS, Linux, and Windows:

claude mcp add --transport http lokeye https://www.lokeye.in/mcp --header "X-API-Key: YOUR_API_KEY"
lokeye uses the Streamable HTTP MCP transport — no stdio process needed. Each request is stateless and authenticated via the X-API-Key header.
Server Capabilities
What the server exposes
The lokeye MCP server implements all three MCP building blocks.

Tools

Functions your model actively calls to dispatch tasks, review submissions, manage credits, and locate humans. The model decides when to call each tool based on your request.

11 tools — model-controlled

Resources

Read-only data your agent can pull in as context — credit balance, task lists, and live human locations. Application-driven: your client decides what to include.

3 resources, 2 templates — application-driven

Prompts

Pre-built instruction templates for common workflows. Invoke them with arguments to guide your model through availability checks, task creation, and area surveys.

3 prompts — user-controlled

Tools
Available tools
11 tools the model can invoke. Discovered via tools/list, executed via tools/call.
create_task
Dispatch a microtask to a nearby human to capture real-world data. Types: photo or video. Bounty in rupees (min ₹10). Tasks go to humans within 250m of the location. Title should be just the subject — e.g. "Phoenix Mall", not "Take a photo of Phoenix Mall".
typerequiredphoto or video
titlerequiredSubject to capture (3-200 chars)
descriptionrequiredWhat is needed (10-2000 chars)
instructionsSpecific instructions for the human
latituderequiredTarget latitude (-90 to 90)
longituderequiredTarget longitude (-180 to 180)
bounty_rupeesrequiredPayment in rupees (10-1000)
deadlinerequiredISO 8601 deadline
get_task_status
Get current status and details of a task. If completed, includes the human response with photo/video URLs.
task_idrequiredTask UUID
list_my_tasks
List your tasks. Optionally filter by status: pending, assigned, in_progress, submitted, completed, expired, cancelled.
statusFilter by status (optional)
cancel_task
Cancel a pending or assigned task. Bounty is refunded to your credits.
task_idrequiredTask UUID
approve_task
Approve a submitted task. Triggers human payout. You have 2 minutes before auto-approval.
task_idrequiredTask UUID
reject_task
Reject a submitted task with a reason. Task returns to the pool. Agents rejecting >35% of submissions may be flagged.
task_idrequiredTask UUID
reasonrequiredRejection reason shown to human (5-500 chars)
view_task_media
Get media URLs from a completed task. Photos are returned inline as base64 images. Media is auto-deleted — photos after 30 min, videos after 5 min. Approve the task before downloading.
task_idrequiredTask UUID
check_credits
Check your current credit balance in rupees.
No parameters
request_credits
Request a credit top-up. You will be contacted via your registered email with payment details. Final payable might be slightly higher to account for conversion/transfer rates.
amount_rupeesrequiredAmount in rupees (₹100 to ₹10,000)
check_human_availability
Count active humans within 250m of a location. Use before creating a task to verify coverage. Minimum INR 10 credits required.
latituderequiredLatitude (-90 to 90)
longituderequiredLongitude (-180 to 180)
get_human_locations
Get anonymized coordinates of all active humans seen in the last 30 minutes. Minimum INR 10 credits required.
No parameters
Resources
Available resources
Read-only context your MCP client can pull in. Discovered via resources/list and resources/templates/list, read via resources/read. All return application/json.

Direct resources

lokeye://credits
Your current credit balance in rupees. Returns { balance, currency, balance_display }.
lokeye://tasks
Your 50 most recent tasks across all statuses, ordered by creation date descending.
lokeye://humans/locations
Anonymized lat/lng coordinates of all active humans seen in the last 30 minutes. Requires INR 10 minimum credits.

Resource templates

lokeye://tasks/{task_id}
Full details for a specific task, including the submitted response and media URLs if available. Replace {task_id} with a task UUID.
lokeye://humans/availability/{lat}/{lng}
Count of active humans within 250m of the given coordinates. Replace {lat} and {lng} with decimal values. Requires INR 10 minimum credits.
Prompts
Available prompts
Pre-built instruction templates. Discovered via prompts/list, retrieved via prompts/get with arguments. MCP clients typically expose these as slash commands.
capture_photo
Guides your model through dispatching a photo task: checks human availability at the target location, then creates the task if coverage is sufficient.
subjectrequiredWhat to photograph, e.g. "Phoenix Mall entrance"
latituderequiredTarget latitude
longituderequiredTarget longitude
bounty_rupeesPayment in rupees (10–1000, default 20)
detailsAdditional instructions for the human
capture_video
Same as capture_photo but for video tasks. Checks availability, then dispatches a video recording task.
subjectrequiredWhat to record, e.g. "Traffic at Main Street junction"
latituderequiredTarget latitude
longituderequiredTarget longitude
bounty_rupeesPayment in rupees (10–1000, default 40)
detailsAdditional instructions for the human
survey_area
Survey a location for human coverage before deciding whether to dispatch a task. Returns a coverage report with human count and overall map of active humans.
latituderequiredTarget latitude
longituderequiredTarget longitude
task_typephoto or video (default: photo)
Example
Typical agent workflow

Check availability

Call check_human_availability (tool) or read lokeye://humans/availability/{lat}/{lng} (resource) to verify humans are nearby before spending credits.

Create task

Call create_task with type, location, bounty, and deadline. Or invoke the capture_photo / capture_video prompt to let the model handle both steps automatically.

Monitor status

Poll get_task_status or read the lokeye://tasks/{task_id} resource. Set up a webhook via the API to get notified instantly on submission.

Review & download

Call view_task_media to review submitted photos/videos (returned as base64). Then call approve_task or reject_task. Download promptly — media is auto-deleted after 30 min (photos) / 5 min (videos).

Set up a webhook URL via the API to get notified instantly when a human submits — no polling needed.
lokeye - The lean human layer for AI