Skip to main content
Glama

dejo_run

Run browser tasks on any website: read values, fill forms, download files, or check pages; replay saved routes or explore new steps.

Instructions

Use for ANY website task: read a value, fill a form, download, check a page. A real browser with memory: a task done once is replayed later with no page reading.

Call with site, task and url. Known task: Dejo replays it and returns the answer. New task: Dejo opens the url and returns the page, one line per element, each with a ref (e12). Then call again with steps, one per line. They run in order. Dejo stops at the first step that fails and returns the page there.

Steps: go URL click T | fill T "text" | select T "option" | check T | press Enter (in a multi-line box, \n in the text starts a new line) read T as name (the value you will report) expect url "part" | expect text "words" | expect see T look words (show the page, these words first; look image: a picture) Also: back, hover T, wait T, upload T "file", tab N|close, dialog dismiss, screenshot [T], pdf, mouse click X Y, click T button=right, read table T, read fields '{"name": "css"}' T, read response "/api", wait response:/api, emulate device "iPhone 13"|timezone X, record video|trace. help (every other step) T is a ref, or role "name" (link "Travel"), text "...", label "...", css "...". Add first, last or nth 3 to mean that place in a list, not that item. Write click! (or fill! etc.) for a step that buys, pays, sends, deletes or submits.

Passed steps are kept. When all steps pass and one is a read (or a file downloads), the task is saved. done=true saves a task that reads nothing.

Tool call template

Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {
      "site": "<site>",
      "task": "<task>"
    },
    "name": "dejo_run"
  }
}

Input Schema

NameRequiredDescriptionDefault
urlNowhere to start
doneNo
siteYesdomain or URL
taskYesthe job in plain words; the trail's name
stepsNostep lines, see above
profileNosigned-in role to use, e.g. admin

Schema Changelog

Changes observed during successful MCP inspections.

  1. First observedv0.5.0

TDQS

A4.2/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations supplied, the description carries the full burden and does it well: it discloses replay-from-memory behavior, that execution halts at the first failing step and returns the page there, that passed steps persist across calls, and the exact conditions under which a task is saved (all steps pass plus a read or download, or done=true). It also flags consequential actions via the 'click!' suffix convention, which is behavior the agent could not otherwise infer.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The purpose is front-loaded in the first sentence, and the step grammar is broken into labelled blocks rather than prose. It is dense and telegraphic ('read T as name (the value you will report)'), which is justified by the DSL surface area, but the terse fragments cost some parseability.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a six-parameter tool with no output schema and a mini step language, the description covers the workflow, the step vocabulary, element reference syntax, list-position modifiers, and task-saving semantics — enough to call it correctly. What it omits is how failures other than a step failure surface, and how profile/auth interacts with saved tasks.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 83%, so the baseline is 3; the description adds real meaning on top by explaining the two-phase use of url vs steps, that task doubles as the trail's name, and that done=true saves a task that reads nothing. profile is the one gap — it is documented only in the schema, with no note on how a signed-in role interacts with replay.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose4/5

Does the description clearly state what the tool does and how it differs from similar tools?

The opening line states a concrete verb and resource: drive a real browser to read a value, fill a form, download or check a page, and it adds the distinctive memory/replay trait ('a task done once is replayed later with no page reading'). That is far more specific than a name restatement, but it never distinguishes itself from siblings like dejo_login or dejo_profile, which an agent could plausibly confuse with 'ANY website task'.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description lays out an explicit call protocol: first call with site/task/url, then call again with steps, and it explains the known-task vs new-task branch. It also defines the escape hatch ('help' returns every other step). It stops short of naming when to prefer a sibling tool, so there are no exclusions.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.