Opencode Workbench

com.ascent-partners由 ascent-partners.com 发布v5.0.1更新于 Oct 29, 2026

Provision an OpenCode workbench and MCP stack on any Linux box, local or over SSH.

发布方自己的服务:Coso Mcp

已验证Streamable HTTP可网页运行云端与基础设施开发者工具

概览

AI 生成的概览

在本地或可通过 SSH 访问的 Linux 机器上部署可复现的 OpenCode 开发工作站:检查、规划、克隆、验证,并管理组件与凭据。

可以做什么
  • 描述内置 Workbench 配置档、组件分层与可选 MCP 附加项
  • 检查本地或 SSH 目标机的系统、工具链、Docker 以及已存在哪些凭据(仅名称)
  • 规划并执行完整配置克隆,随后逐组件验证结果
  • 安装、更新或移除单个组件,如 OpenCode、pnpm、MCP 服务器、技能与 Docker 容器
  • 列出所需凭据,并为缺失项给出获取指引
适用场景
当你希望在全新或现有 Linux 机器上(本地或通过 SSH)重建或统一 OpenCode 环境及其 MCP 服务器、技能和插件,并希望先看到只读计划再动手时使用。
运行要求
可作为远程 streamable HTTP 端点使用,也可通过 npx 从 npm 包本地运行(需要 Node.js)。目标机须为 Linux,可本地访问或通过 SSH 以用户凭据访问。配置档引用多个服务商密钥(如 DEEPSEEK_API_KEY、OPENCODE_API_KEY、GitHub PAT、PERPLEXITY_API_KEY、BROWSERLESS_TOKEN),服务器只检查是否已设置。
安装前请注意
apply_clone、install_component 和 update_component 会写入目标机:覆盖 OpenCode 配置文件、复制 AGENTS.md、将已有配置仓库重置到 origin/main,并执行全局安装。apply_clone 需 confirm:true 才执行。可能使用 SSH 访问与特权命令。README 称从不读取或传输密钥值、只写入空的 env 模板,但目标机器及其配置会被修改。

安装

在 SourceWeft 中

  1. 打开 控制台中的 Opencode Workbench,将其添加到工作区。
  2. 为需要使用其工具的对话启用该服务。

Web executable,通过 Streamable HTTP。 远程服务在工作区中配置后即可从网页运行时运行。

交给智能体安装

把这段提示词粘贴到 Claude Code、Codex、Cursor 等智能体,或 SourceWeft 对话中。智能体会读取该服务器的安装指南,先向你展示发布者、端点和登录方式,经你同意后再把它添加到你的客户端。密钥不会写进任何命令。

阅读 https://sourceweft.com/mcp/com.ascent-partners%2Fopencode-workbench/install.md,并按其中说明把这个 MCP 服务器接入我的客户端。修改任何配置之前,先告诉我它的发布者、端点和登录方式,等我确认。不要把 API 密钥写进命令或我会提交的文件;问我密钥应从哪里获取。

查看智能体所依据的安装指南

其他 MCP 客户端

把它添加到你客户端的 mcpServers 配置中。

{
  "mcpServers": {
    "opencode-workbench": {
      "type": "http",
      "url": "https://opencode-workbench.simonmak.com/mcp"
    }
  }
}

README

OpenCode Workbench

A reproducible, self-documenting, and recoverable OpenCode development workstation — configuration, agent skills, MCP stack, and an MCP server that clones the entire setup onto any Linux machine (locally or over SSH).

[Secret Scan] [MCP Tool Definition Quality] [MCP Registry] [Glama] [MCP Server] [Validate Documentation] [License: MIT]

Homepage & remote MCP: https://opencode-workbench.simonmak.com · POST https://opencode-workbench.simonmak.com/mcp

Overview

This repository is the single source of truth for a complete cloud development workstation. Everything needed to rebuild from scratch is versioned here:

  • Workstation configuration and automation
  • OpenCode configuration (model, providers, MCP servers)
  • Agent skills (76 reusable AI instructions across research, dev, data, science and ops)
  • Operational prompts (disaster recovery, backup, security)
  • Recovery procedures (playbook, checklist, gap analysis)
  • Security governance (secret management, threat model)
  • OpenCode sandbox (bubblewrap) — run the agent with a read-only system and no secret mounts
  • opencode-workbench — an MCP server that installs this profile on any Linux box

No important knowledge exists only in human memory.

Clone This Workbench to Another Linux Machine

Register the MCP server with OpenCode:

json
{  "mcp": {    "opencode-workbench": {      "type": "local",      "command": ["npx", "-y", "@simonmak-ascent/opencode-workbench"],      "enabled": true,      "timeout": 600000    }  }}

Then ask the agent to run inspect_target → plan_clone → apply_clone { confirm: true } → verify_clone, for { "mode": "local" } or { "mode": "ssh", "host": "<your-box>", "user": "<your-user>" }. apply_clone is consent-gated: without confirm:true it returns a plan (components + privileged-command preview) and changes nothing. Run list_required_credentials to see which keys the box still needs and how to acquire them, and bash scripts/selftest/run-selftest.sh to verify it works. See mcp-server/README.md and docs/architecture/clone-mcp.md.

The clone is key-resilient: with no model key it still boots on the OpenCode Zen free floor, and MCP servers whose credentials are absent are disabled and reported (not left to fail at runtime).

The connector never reads or transmits secret values; it writes an empty ~/.env.workbench template for you to fill in.

Quick Start (≤ 5 minutes)

Fastest path: no box required — register the MCP server with npx -y @simonmak-ascent/opencode-workbench (config in Clone This Workbench); or provision a fresh box:

bash
# Create a new build boxgh SWAS create --repo simonmak-ascent/opencode-workbench --machine basicLinux32gb
# Wait for postCreate (3-5 min) — installs everything automatically# Then start coding:opencode

Architecture

mermaid
flowchart TB  AGENT["OpenCode agent"] -->|"npx @simonmak-ascent/opencode-workbench"| MCP["opencode-workbench MCP<br/>inspect · plan · apply · verify"]  REPO[("Git repo — single source of truth")] --> MCP  MCP -->|"local or SSH"| BOX["Target Linux box"]  BOX --> PROF["Workstation profile<br/>OpenCode CLI · 31 MCP servers · 76 skills · Docker infra"]

Clone lifecycle

apply_clone is consent-gated: without confirm: true it returns a plan and changes nothing.

mermaid
sequenceDiagram    autonumber    participant A as OpenCode agent    participant M as opencode-workbench MCP    participant B as Target box (local or SSH)    A->>M: inspect_target    M-->>A: OS, toolchain, Docker, existing config    A->>M: plan_clone    M-->>A: components + privileged-command preview    A->>M: apply_clone { confirm: true }    M->>B: install components · write ~/.env.workbench template    B-->>M: result    A->>M: verify_clone    M-->>A: pass/fail + missing credentials

MCP tool surface

mermaid
flowchart LR    A["OpenCode agent"] --> D["Discovery<br/>describe_workbench · inspect_target"]    A --> C["Clone<br/>plan_clone · apply_clone · verify_clone"]    A --> K["Components<br/>install · remove · update_component"]    A --> S["Credentials<br/>list_required_credentials · run_auth_flow"]
opencode-workbench/├── mcp-server/                     # opencode-workbench: clone this profile onto Linux│   ├── src/                        # inspect / plan / apply / verify, local + SSH│   ├── test/                       # unit tests (vitest)│   └── README.md├── connector.json                  # MCP connector manifest / registration snippet├── plugins/                        # memory.ts + doc-tools.ts OpenCode plugins├── package.json                    # thin launcher for opencode-workbench (bin + prepare)├── .devcontainer/                  # build box definition & bootstrap│   ├── devcontainer.json           # Container image, features, ports, remoteEnv│   ├── setup.sh                    # Post-create: installs opencode, MCPs, Docker infra│   └── aliases.sh                  # Shell aliases sourced at runtime├── .opencode/│   └── skills/                     # 76 agent skills, one directory per SKILL.md├── vendor/                         # Vendored MCP source code│   ├── perplexity-agent-mcp/│   └── browserless-mcp/├── configs/                        # Reference configurations│   ├── mcp/                        # MCP server inventory & architecture│   ├── opencode/                   # OpenCode settings documentation│   ├── shell/                      # Shell configuration│   ├── git/                        # Git configuration & hooks│   └── providers/                  # AI provider configuration├── docs/                           # Complete documentation system│   ├── WORKBENCH_CHARTER.md        # Governance: mission, principles, requirements│   ├── security-model.md           # Security architecture & threat model│   ├── IMPROVEMENT_BACKLOG.md      # Tracked improvements and fixes│   ├── CHANGELOG_WORKBENCH.md      # Append-only change history│   ├── architecture/               # System documentation (incl. clone-mcp.md)│   ├── recovery/                   # Recovery procedures (4 docs)│   ├── prompts/                    # Operational prompt library (9 prompts)│   └── backup-reports/             # Dated backup snapshots├── scripts/                        # Automation scripts│   ├── backup/                     # Master backup script│   ├── inventory/                  # Auto-generate inventories (software, env, MCP, OpenCode)│   ├── security/                   # Secret pattern scanner│   ├── recovery/                   # Recovery validation│   └── maintenance/                # Config sync, routine tasks├── .github/workflows/              # CI (docs, inventory, gitleaks, mcp-server)├── opencode.json                   # OpenCode configuration (MCP servers)├── AGENTS.md                       # OpenCode agent instructions└── README.md                       # This file

What's Inside

CategoryDetails
OpenCodeOpenCode 2 CLI, installed from the v2 channel (latest 2.x at clone time; inspect_target reports the version it found) · DeepSeek V4 Pro model · LSP enabled
MCP Servers31 configured (29 enabled, 2 disabled), 8 remote + 23 local
Agent Skills76 reusable skills in .opencode/skills/
Versionsworkbench.versions.json — every MCP server, plugin and skill resolved from its real upstream, regenerated and CI-checked
InfrastructurePostgreSQL 16 (Docker), Browserless Chromium (Docker)
Toolspandoc, jq, miller, sqlite3, GitHub CLI, Playwright Chromium
Prompts9 operational prompts in docs/prompts/

Workstation Governance

This workstation operates under a formal charter. Key principles:

  • Single Source of Truth: Runtime must match repository. Drift is a bug.
  • Secrets Never Touch Disk: All secrets flow through the build box Secrets.
  • Immutable History: Changelog is append-only. Every change is traceable.
  • Documentation Lives With Code: Docs describe the system as it is, not as imagined.

Read the full charter: docs/WORKBENCH_CHARTER.md

Backup Workflow

bash
# Master backup — full inventory, security scan, commit prepbash scripts/backup/run-master-backup.sh
# Quick backup — save state before risky operationsbash scripts/maintenance/sync-runtime-config.sh && \  git diff > /tmp/quick-backup-$(date +%Y%m%d).diff

Or use OpenCode prompts:

  • docs/prompts/RUN_MASTER_BACKUP.md — execute backup via AI
  • docs/prompts/QUICK_BACKUP.md — rapid state preservation

Recovery Workflow

Assume the build box is deleted. Only Git repo + build box Secrets survive.

bash
# 1. Create new build boxgh SWAS create --repo simonmak-ascent/opencode-workbench --machine basicLinux32gb
# 2. Wait for automation (3-5 min)
# 3. If needed (usually done by setup.sh):npx playwright install chromium# Vercel MCP uses VERCEL_ACCESS_TOKEN (token auth — no OAuth step)
# 4. Validate recoverybash scripts/recovery/validate-recovery.sh

Full details: docs/recovery/RECOVERY_PLAYBOOK.md

MCP Architecture

TypeCountNotes
Remote9context7, gh_grep, vdd, exa, cloudflare, clerk, vercel (token auth), sentry, stripe
Local26npm / vendored / binary stdio servers

Totals: 31 configured (29 enabled, 2 disabled: google-search, google-workspace).

See: docs/architecture/mcp-inventory.md

Secrets Management

All secrets stored in the build box Secrets. Never in repository files.

#SecretUsed By
1DEEPSEEK_API_KEYOpenCode (primary model)
2OPENCODE_API_KEYOpenCode Zen / Console Go providers
3SIMONMAK_ASCENT_PATGitHub MCP, shadcn MCP, gh
4PERPLEXITY_API_KEYPerplexity MCP
5BRAVE_API_KEYBrave Search MCP
6BROWSERLESS_TOKENBrowserless MCP + container
7SENTRY_AUTH_TOKENSentry MCP
8VERCEL_ACCESS_TOKENVercel MCP (token auth)

Additional keys for opt-in servers (EXA_API_KEY, CLOUDFLARE_API_TOKEN, STRIPE_SECRET_KEY, SURREAL_*, ALIBABA_CLOUD_*, AZURE_*, FIRECRAWL_API_KEY, FRED_API_KEY, COMPANIES_HOUSE_API_KEY) are catalogued in docs/architecture/secrets.md — never by value.

See: docs/architecture/secrets.md

Documentation Index

DocumentPurpose
docs/WORKBENCH_CHARTER.mdMission, principles, governance
docs/security-model.mdSecurity architecture, threat model, incident response
docs/IMPROVEMENT_BACKLOG.mdTracked improvements and technical debt
docs/CHANGELOG_WORKBENCH.mdComplete change history
docs/architecture/System documentation (10 inventory docs)
docs/recovery/Recovery playbook, disaster recovery, checklist, gap analysis
docs/prompts/9 operational prompts with index
docs/backup-reports/Dated backup snapshots

Adding to the Workbench

  1. MCP Server: Add npm package to .devcontainer/setup.sh, add to opencode.json, document in docs/architecture/mcp-inventory.md
  2. Agent Skill: Create .opencode/skills/<name>/SKILL.md, follow naming convention
  3. Tool: Add to scripts/bootstrap-tools.sh
  4. Prompt: Create in docs/prompts/, update prompt-index.md
  5. After any change: Update CHANGELOG_WORKBENCH.md, run master backup

Recovery Score: 94/100

Validated disaster recovery within < 10 minutes using repository + build box Secrets alone.

Use with Context7

Up-to-date OpenCode Workbench documentation is indexed on Context7, so coding agents can pull it into context on demand. With the Context7 MCP server or ctx7 CLI installed, name the library in your prompt:

text
use library /simonmak-ascent/opencode-workbench for API and docs

License

MIT — see LICENSE.


A tool by Simon Mak.

If this saves you time, a ⭐ on GitHub helps others find it.

来源:README.md,提交 976778f

工具

10

describe_workbench

Workbench server info

Describe this MCP server from its bundled profile without contacting any target or making a network call: repository URL, the default component set, components grouped by tier (required/core/optional), and optional MCP add-ons. It takes no parameters, is always safe, and returns instantly. Use it first to resolve a component id for install_component/remove_component/update_component or to see available add-ons; use inspect_target for a machine's live state.

无需输入参数。

inspect_target

Inspect a Linux target

Probe one Linux machine — this host (`local`) or a remote host over SSH (`ssh`) — and report its OS, kernel, architecture, package manager, Node/npm, OpenCode, Docker, and per-tool detection flags, plus the names (never values) of credentials present. Read-only: it runs one shell probe as the target user (over SSH when `mode:ssh`), installs nothing, writes nothing, contacts no external service, and can take a few seconds per hop. Parameter semantics: `target.mode` selects local vs ssh (default `local`); `target.host` is required only for ssh; `target.user` defaults to the ssh config/current user; `target.cwd` sets the directory for relative checks; `target.port`/`identityFile` pass straight to ssh. Use it to see what a clone would touch, then plan_clone to turn that into an ordered plan and apply_clone to perform it.

输入参数target必填

plan_clone

Plan a Workbench clone

Return a read-only plan for a target: compare it against the Workbench profile and list each component as `install`, `present`, or `manual`, with the commands that would run. It still runs a detection probe on the target (same access as inspect_target) but clones nothing and writes nothing. Parameter semantics: `components` restricts the diff to those ids (default: required+core); `workspace` sets the expected profile path on the target (default `~/opencode-workbench`); `profileUrl` overrides the git remote. Use it to preview before apply_clone, which performs the changes.

输入参数dryRuntarget必填skipRepoworkspacecomponentsprofileUrl

apply_clone

Apply a Workbench clone

Apply the Workbench profile to a target: clone the repo and install missing components (OpenCode CLI, Node/pnpm, MCP servers, skills, plugins, optional Docker). It OVERWRITES `<home>/.config/opencode/opencode.json`, copies `AGENTS.md`, resets an existing profile repo to `origin/main`, and runs global installs that need write permission (plus SSH for `mode:ssh`) and can take minutes. Writes a names-only `~/.env.workbench` (mode 600); never secret values. Consent-gated and idempotent: without `confirm:true` it returns the plan and changes nothing. Parameter semantics: `components` defaults to required+core; `workspace` defaults to `~/opencode-workbench`; `skipRepo` skips the clone/update; `profileUrl` overrides the git remote; `dryRun` returns the plan. To preview only, use plan_clone.

输入参数dryRuntarget必填confirmskipRepoworkspacecomponentsprofileUrl

verify_clone

Verify a Workbench clone

Re-check a target after a clone: return a per-component `present`/`missing` list plus whether `opencode.json` and `~/.env.workbench` exist. Read-only and idempotent: it runs a detection probe (SSH or local) and checks files, writing nothing. `target` selects the machine (`local` or SSH) and its `home` is where the `opencode.json` and `~/.env.workbench` checks run; `components` restricts the check to those ids (omit it to check every component, the default); `workspace` sets the profile path used by component checks. Use it after apply_clone and after component changes; for a pre-clone preview use plan_clone.

输入参数dryRuntarget必填skipRepoworkspacecomponentsprofileUrl

install_component

Install one Workbench component

Install a single Workbench component by id (e.g. `node`, `opencode`, `npm-mcps`, `skills`) on a target; `component` must be an id from describe_workbench. Overwrites that component's files when present (e.g. `skills`/`plugins` replace the copies under `<home>/.config/opencode`); global installs need write permission and can take minutes. Idempotent and consent-gated: without `confirm:true` it returns the plan. Parameter semantics: `component` is required; `workspace` defaults to `~/opencode-workbench`; `confirm` defaults to false (plan only). Use it for one targeted component; use apply_clone to install the full default set.

输入参数target必填confirmcomponent必填workspace

remove_component

Remove one Workbench component

Uninstall one Workbench component by id from a target — the inverse of install_component — for a bounded, documented subset with an automated uninstall (e.g. `opencode`, `pnpm`, `npm-mcps`, `skills`, `docker-containers`); `component` must be an id from describe_workbench. Idempotent: an already-absent component reports `absent`; a component with no automated uninstall (system packages such as git/curl/node) reports `manual`. Destructive and consent-gated: without `confirm:true` it returns a preview and changes nothing, and it deletes only that component's files — never `opencode.json` or the profile repo. Parameter semantics: `workspace` sets the profile path some removals need; `dryRun` reports without changing. Use it to tear down one component; for several, call it per id.

输入参数dryRuntarget必填confirmcomponent必填workspace

update_component

Update one Workbench component

Update one Workbench component in place by id: re-run its install script to fetch the current version (e.g. `opencode` re-runs the official installer; `npm-mcps` reinstalls the latest globals). It OVERWRITES the component's files, needs write permission, and can take minutes for global installs; the resolved version may change. Parameter semantics: `component` must be an id from describe_workbench; `workspace` defaults to `~/opencode-workbench`; `dryRun` previews; `confirm` defaults to false — set `confirm:true` to apply. For the whole profile use apply_clone.

输入参数dryRuntarget必填confirmcomponent必填workspace

list_required_credentials

List required credentials

List the credentials the profile references, which the target already provides, and how to acquire each missing one. Value-blind: it reads only whether each env var is set — no values, no network, no writes — via one read-only probe. Parameter semantics: `target` selects the machine (`local` or SSH) and its `home` sets the returned `template` path and the env file the presence check reads; there are no other parameters. Returns `present`/`missing` var-name arrays plus one guidance record per credential (`var`, `purpose`, `url`, `method`, `command`). Use it after plan_clone to see what would degrade; act on one credential with run_auth_flow.

输入参数target必填

run_auth_flow

Acquire one credential (best-effort)

Return the acquisition plan for one credential on a target: the provider URL, the exact non-interactive command when one exists (e.g. `opencode auth login`, `opencode mcp auth vercel`, `gh auth login`), and whether it is already present. Read-only and value-blind: it emits guidance and re-checks presence with one read-only probe; it does not run the commands or handle secret values. Parameter semantics: `var` must be an env var name returned by list_required_credentials (unknown names error); `target` selects the machine and its `home` is where the value should be written. Use it for a single credential; it does not replace that listing.

输入参数var必填target必填

版本历史

2
  1. v5.0.1最新Oct 29, 2026
  2. v4.0.2Oct 14, 2026

评分

已满足适用检查项的 100%

该评分衡量此收录条目的完整程度,依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价,也不表示服务器安全或当前可用:可连通性是带日期的检查结果,评分也不包含安装量或用户评价。

  • 近 30 天内可连通必需25 分
  • 已声明认证方式必需10 分
  • 已列出工具必需15 分
  • 描述必需15 分
  • 图标10 分
  • 文档15 分
  • 许可证10 分

其他 Ascent-Partners 集成