Colibri

Esegui passi

browse_do

Run up to 40 ordered browser actions, return only the changed content diff, and stop at the first error or block.

Instructions

Run up to 40 steps in order, return only the diff; stops at the first error or block. view: diff|full|none.

Input Schema

Name	Required	Description
`view`	No
`steps`	Yes	One verb per step. T = ref "e12" \| name \| {name,role,row,n} \| {x,y} 0-1 (canvas/maps). click:T (force:true if covered) · fill:T+text (date YYYY-MM-DD) · type:T+text (keystrokes) · select:T+option (label/value/part) · check:T (on:false) · press:key (+target) · pick:"text" (visible option) · hover:T · scroll:down\|up\|top\|bottom\|T\|px · wait:ms\|"idle" · waitFor:"visible text" (+timeout) · goto:url · back:true · tab:n\|"list"\|"close" · upload:T+file (T may be a drop area) · dialog:accept\|dismiss · find:"text" · read:text\|table\|cards\|fields\|describe (+target, max, links, sort, filter, maxPrice, query as browse_read) · confirm:true only after the user OKs a !c step.
`script`	No	Async JS body run after steps in a local sandbox: colibri.open\|search\|view\|read\|do\|task\|stats(args) as the tools (same safety), colibri.rows, text(), tables(). Return only the result (max 6k tokens).

Schema Changelog

Changes observed during successful MCP inspections.

Changed2 schema fields changedv0.5.11Nov 5, 2026

addedInput schema / properties / script

Added value: +{
+  "description": "Async JS body run after steps in a local sandbox: colibri.open|search|view|read|do|task|stats(args) as the tools (same safety), colibri.rows, text(), tables(). Return only the result (max 6k tokens).",
+  "type": "string"
+}

changedInput schema / properties / steps / description

Previous value: -"One verb per step. T = ref \"e12\" | name text | {name,role,row,n} | {x,y} 0-1 for canvas/maps. click:T (force:true = JS click if covered) · fill:T+text (native date: YYYY-MM-DD) · type:T+text (keystrokes, autocomplete) · select:T+option (native <select>; label, value or part) · check:T (on:false unchecks) · press:key (+target:T) · pick:\"text\" (visible suggestion/option) · hover:T · scroll:down|up|top|bottom|T · wait:ms (max 10000)|\"idle\" (network+spinners done) · waitFor:\"visible text\" (+timeout ms) · goto:url · back:true · tab:n|\"list\"|\"close\" · upload:T+file (path or list; T may be a \"drag here\" area) · dialog:accept|dismiss (next alert/confirm) · find:\"text\" (matching lines) · read:text|table|cards (+target, max, links) · confirm:true only after the user OKs a !c step."New value: +"One verb per step. T = ref \"e12\" | name | {name,role,row,n} | {x,y} 0-1 (canvas/maps). click:T (force:true if covered) · fill:T+text (date YYYY-MM-DD) · type:T+text (keystrokes) · select:T+option (label/value/part) · check:T (on:false) · press:key (+target) · pick:\"text\" (visible option) · hover:T · scroll:down|up|top|bottom|T|px · wait:ms|\"idle\" · waitFor:\"visible text\" (+timeout) · goto:url · back:true · tab:n|\"list\"|\"close\" · upload:T+file (T may be a drop area) · dialog:accept|dismiss · find:\"text\" · read:text|table|cards|fields|describe (+target, max, links, sort, filter, maxPrice, query as browse_read) · confirm:true only after the user OKs a !c step."

Changed1 schema field changedv0.5.8Oct 31, 2026

changedInput schema / properties / steps / description

Previous value: -"One verb per step. T = ref \"e12\" | name text | {name,role,row,n} | {x,y} 0-1 for canvas/maps. click:T (force:true = JS click if covered) · fill:T+text (native date: YYYY-MM-DD) · type:T+text (keystrokes, autocomplete) · select:T+option (native <select>; label, value or part) · check:T (on:false unchecks) · press:key (+target:T) · pick:\"text\" (visible suggestion/option) · hover:T · scroll:down|up|top|bottom|T · wait:ms (max 10000) · waitFor:\"visible text\" (+timeout ms) · goto:url · back:true · tab:n|\"list\"|\"close\" · upload:T+file (path or list) · dialog:accept|dismiss (next alert/confirm) · find:\"text\" (matching lines) · read:text|table (+target, max) · confirm:true only after the user OKs a !c step."New value: +"One verb per step. T = ref \"e12\" | name text | {name,role,row,n} | {x,y} 0-1 for canvas/maps. click:T (force:true = JS click if covered) · fill:T+text (native date: YYYY-MM-DD) · type:T+text (keystrokes, autocomplete) · select:T+option (native <select>; label, value or part) · check:T (on:false unchecks) · press:key (+target:T) · pick:\"text\" (visible suggestion/option) · hover:T · scroll:down|up|top|bottom|T · wait:ms (max 10000)|\"idle\" (network+spinners done) · waitFor:\"visible text\" (+timeout ms) · goto:url · back:true · tab:n|\"list\"|\"close\" · upload:T+file (path or list; T may be a \"drag here\" area) · dialog:accept|dismiss (next alert/confirm) · find:\"text\" (matching lines) · read:text|table|cards (+target, max, links) · confirm:true only after the user OKs a !c step."

First observedv0.4.0Oct 29, 2026

TDQS

A3.5/5.0

Behavior3/5

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

With no annotations, the description carries the full disclosure burden. It does contribute real behavior: a hard cap of 40 steps, in-order execution, diff-only return, and fail-fast on the first error or block. It omits that these steps perform real browser mutations (click/fill/upload/dialog), whether actions are reversible, and what the diff is measured against.

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?

Two short, front-loaded sentences with no filler; the constraint and failure semantics come first. The telegraphic style ('view: diff|full|none') is efficient but leans on the reader already knowing what those modes mean.

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

Completeness3/5

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

For a three-parameter mutation tool with no annotations and no output schema, the description should explain side effects and the return shape more fully. Limits and failure behavior are covered, but the safety profile and the structure of the returned diff are largely left to the schema.

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 67% and the 'steps' grammar is exhaustively documented in the schema. The description adds value on 'view' by pairing it with 'return only the diff', which reveals diff as the default and the trade-off against full/none. It does not define what 'full' or 'none' actually return, nor how 'script' composes with 'steps'.

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?

States a specific action and resource: 'Run up to 40 steps in order' with a bounded sequence and a stated return shape. It never names a sibling, so an agent must infer that this is the execution tool versus browse_open/browse_task/browse_read, but the verb-plus-resource phrase is concrete enough to be actionable.

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

Usage Guidelines3/5

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

The only usage signal is implicit: this is for executing a step sequence, and the schema ties 'read' steps back to browse_read and 'confirm' to a user approval gate. There is no explicit when-to-use versus browse_task or browse_open, and no stated prerequisites or exclusions.

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

Deploy Server

Esegui passi

Instructions

Input Schema

Schema Changelog

TDQS

Other Tools