opencode-workbench

Apply a Workbench clone

apply_clone

DestructiveIdempotent

Install the Workbench profile on a target: clone the profile repo and install missing components (repo, OpenCode CLI, Node/pnpm, npm/vendored/research MCPs, skills, plugins, optional Docker), writing a rendered opencode.json and an empty, names-only ~/.env.workbench (mode 600). Idempotent — present components are skipped. Consent-gated: without confirm:true it returns the plan and changes nothing. Use it for a component subset or per-component control on an already-provisioned host; for a first-time provision use bootstrap_host, and do not call both for the same host and change.

Input Schema

TableJSON Schema

Name	Required	Description
`dryRun`	No	Report only; make no changes.
`target`	Yes	The machine to operate on: this host (local) or a remote host over SSH (ssh).
`confirm`	No	Set true to actually install. When absent, the call returns a plan and makes no changes.
`skipRepo`	No	Do not clone/update the profile repo on the target.
`workspace`	No	Target directory for the profile repo.
`components`	No	Component ids to include; defaults to required+core.
`profileUrl`	No	Override profile git URL.

Output Schema

TableJSON Schema

Name	Required	Description
`plan`	No	Presented when confirmation is required.
`model`	Yes	Default model selected for the target (DeepSeek when its key is present, else the OpenCode Zen free floor).
`dryRun`	Yes	True when the run made no changes.
`target`	Yes	Label of the target.
`envPath`	Yes	Path of the written ~/.env.workbench template.
`envVars`	Yes	Environment variable names listed in the env template.
`skipped`	Yes	Component ids skipped (already present or manual).
`degraded`	Yes	Capabilities disabled because required credentials are absent; fill the named env vars to enable them.
`installed`	Yes	Components installed, with exit codes and output.
`workspace`	Yes	Directory where the profile repo was cloned.
`configPath`	Yes	Path of the written opencode.json.
`providerMode`	Yes	Which provider the model resolves to.
`requiresConfirmation`	No	True when the call returned a plan without applying; re-call with confirm:true to install.

Schema Changelog

Changes observed during successful MCP inspections.

First observedOct 2, 2026

TDQS

A4.6/5.0

Behavior4/5

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

Annotations already declare idempotentHint and destructiveHint, but the description adds real context: the consent gate requires confirm:true or nothing changes, present components are skipped, and it writes specific files with mode 600. It stops short of describing permissions/privilege requirements or rollback behavior for a mutation tool, so it is strong but not exhaustive.

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?

Dense but front-loaded: the main action and its side effects come first, then the idempotency and consent caveats, then the routing rule. The long inline component list is information-dense rather than padding, though it does make the single paragraph heavy.

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

Completeness5/5

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

An output schema exists, so return values need no explanation, and the description still covers the mutation semantics, the consent gate, idempotency, the files produced, and the alternative tool. Nothing an agent needs to invoke it correctly is missing.

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 100%, so baseline is 3, but the description goes beyond it by explaining the semantics of the consent gate (returns a plan and no changes when confirm is absent) and by framing the components array as selecting a subset with per-component control. dryRun and the SSH target fields are left to the schema.

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

Purpose5/5

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

It names a specific verb+resource (install the Workbench profile on a target by cloning and installing enumerated components) and enumerates the exact artifacts it produces (rendered opencode.json, mode-600 ~/.env.workbench). This differentiates it clearly from siblings like bootstrap_host and plan_clone without opening any schema.

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

Usage Guidelines5/5

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

Explicit routing: use this for a component subset or per-component control on an already-provisioned host; use bootstrap_host for a first-time provision; and do not call both for the same host and change. Both the when-to-use and the when-not-to-use alternative are stated.

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

All opencode-workbench tools Inspect

Try in Browser

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Glama MCP Gateway

Add one secure layer between your agents and this server.