Atlas / MCP servers / madappgang / claudish

claudishBLOCK

Name: claudish
Author: madappgang

mcp/madappgang/claudish

MCP server by madappgang · ai · claude · claudecode · Audited 2026-10-12

⚑ Report / re-scan Source ↗

Claude Code. Any Model. The most powerful AI coding agent now speaks every language.

Verdict

BLOCK

Grade

Trust score

31 /100

Exposed tools

63 52r · 11w · 0d

Transport

stdio

License

—

Stars

1,005

Overview

From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.

Claude Code. Any Model.

[](https://www.npmjs.com/package/claudish) [](LICENSE) [](https://claude.ai/claude-code)

Use your existing AI subscriptions with Claude Code. Works with Anthropic Max, Gemini Advanced, ChatGPT Plus/Codex, Kimi, GLM, OllamaCloud — plus 580+ models via OpenRouter and local models for complete privacy.

Website · Documentation · Report Bug

Claudish (Claude-ish) is a CLI tool that allows you to run Claude Code with any AI model by proxying requests through a local Anthropic API-compatible server.

Supported Providers:

Cloud: OpenRouter (580+ models), Google Gemini, OpenAI, MiniMax, Kimi, GLM, Z.AI, Sakana Fugu, OllamaCloud, OpenCode Zen
Local: Ollama, LM Studio, vLLM, MLX
Enterprise: Vertex AI (Google Cloud)

Use Your Existing AI Subscriptions

Stop paying for multiple AI subscriptions. Claudish lets you use subscriptions you already have with Claude Code's powerful interface:

Read from source at commit 86d24a12d744OBSERVED · 2026-10-12

Connect

Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.

claude-code

claude mcp add magmux-linux-x64 --env ANTHROPIC_API_KEY=${ANTHROPIC_API_KEY} --env ANTHROPIC_AUTH_TOKEN=${ANTHROPIC_AUTH_TOKEN} --env CLAUDE_CODE_MAX_CONTEXT_TOKENS=${CLAUDE_CODE_MAX_CONTEXT_TOKENS} --env CLAUDE_CODE_OAUTH_TOKEN=${CLAUDE_CODE_OAUTH_TOKEN} -- npx -y @claudish/[email protected]

claude-desktop

{
  "mcpServers": {
    "magmux-linux-x64": {
      "command": "npx",
      "args": [
        "-y",
        "@claudish/[email protected]"
      ],
      "env": {
        "ANTHROPIC_API_KEY": "${ANTHROPIC_API_KEY}",
        "ANTHROPIC_AUTH_TOKEN": "${ANTHROPIC_AUTH_TOKEN}",
        "CLAUDE_CODE_MAX_CONTEXT_TOKENS": "${CLAUDE_CODE_MAX_CONTEXT_TOKENS}",
        "CLAUDE_CODE_OAUTH_TOKEN": "${CLAUDE_CODE_OAUTH_TOKEN}"
      }
    }
  }
}

Exposed tools (63)

52 read · 11 write · 0 destructive.

Tool	Risk	Description
`Artifact`	read	Query a collection.
`Bash`	write	Run a shell command.
`Edit`	write	Edit a file
`ExitPlanMode`	read	original
`GPT-5`	read
`Haiku`	read	Fast & cheap, used for simple tasks
`Mistral`	read	Direct API
`OllamaCloud`	read	Cloud models
`Opus`	read	Most capable, used for complex reasoning
`OtherTool`	read	other
`Read`	read	d
`Sonnet`	read	Balanced, used for general tasks
`Subagent`	read	Used for spawned sub-agents
`TargetTool`	read	claude
`TaskOutput`	read	original TaskOutput description
`WebSearch`	read	original WebSearch description
`Write`	write	Write a file
`Z.AI`	read	Direct API
`advisor`	read	startup probe
`cancel_session`	read	Cancel a session. Returns once its state is CANCELLED ({session_id, state, changed}); the
`capture_session`	read	Read a session
`channel_sleep`	read	Sleeps ${SLEEP_MS}ms emitting notifications/claude/channel every ${TICK_MS}ms, then returns
`compare_models`	write	Run the same prompt through multiple models and compare responses
`create_session`	write	Create a new claudish proxy session for an external model: an interactive Claude Code running
`default`	read	Default profile - shows model selector when no model specified
`first-skill`	read	The real listed skill
`get_diagnostics`	read	Explain what a channel session actually did — its final screen, upstream error bodies,
`get_output`	read	Get a session
`get_weather`	read	Get current weather for a city
`imagegen`	read	Generate images
`kimi-k3`	read	a model
`list_files`	read	List files in a directory
`list_models`	read	List recommended models for coding tasks
`list_sessions`	read	List channel sessions: {contract_version, capabilities, sessions}. Optionally include
`llama`	read
`m`	read
`mcp__browser-use__browser_click`	read	Click an element
`mcp__claudish__run_prompt`	write	Read a prompt and run it
`mcp__plugin_claudish__list_models`	read	List recommended models
`preflight`	read	DIAGNOSTIC. For a list of models, report which provider would serve each,
`probe-log.jsonl`	write	Full diagnostic log for this probe run
`probe_elicit`	read	Asks the human a question via MCP elicitation and reports their answer.
`probe_ping`	read	Returns
`probe_progress`	read	Emits 3 notifications/progress over ~4s then returns. Reports whether the
`probe_resource_link`	read	Returns a resource_link content block pointing at a real file.
`probe_resource_uri`	read	Returns a resource_link using a CUSTOM SCHEME uri (claudish://...) rather than file://.
`probe_structured`	read	Returns structuredContent validated against an outputSchema.
`progress_sleep`	read	Sleeps ${SLEEP_MS}ms emitting notifications/progress every ${TICK_MS}ms, then returns
`query_rows`	read	Query rows from a collection. Each where clause is [field, operator, value].
`read_file`	read	Read a file from disk
`release-notes-writer`	read	Draft release notes
`report_error`	read	Report a claudish error to developers. IMPORTANT: Ask the user for consent BEFORE calling this tool. Show them what data will be sent (sanitized). All data is anonymized: API keys, user paths, and emails are stripped. Set auto_send=true to suggest the user enables automatic future reporting.
`run_prompt`	write	Run a prompt through any model — supports all providers (Kimi, GLM, Qwen, MiniMax, Gemini, GPT, Grok, etc.) with auto-routing, fallback chains, and custom routing rules.
`search_models`	read	Search OpenRouter
`send_input`	write	Send a prompt to a session. Accepted in every non-terminal state and queued until the
`silent_sleep`	read	Sleeps ${SLEEP_MS}ms sending nothing at all, then returns
`simple_ping`	read	Returns
`slow_ping_with_progress`	read	Emits 2 notifications/progress then returns. Used to test whether
`slow_with_many_progress`	read	Runs for ~10 seconds, emitting 5 distinct notifications/progress along the way.
`start_mock_session`	write	Start a mock Claudish session that emits a scripted sequence of channel notifications
`team`	write	Run AI models on a task with anonymized outputs and optional blind judging. Every
`team-session-42-status`	read	Diagnostic status for team session 42
`test_tool`	read	A test tool.

Trust audit

BLOCKgrade F · trust 31/100 Do not install this without reading the findings. The audit found something that could harm you or your machine.

Layer	What it checks	Result
L0	Provenance & inventory	WARN
L1	Static analysis of the code	WARN
L2	Instruction surface (what it tells the agent)	FAIL
L3	Class-specific surface	PASS
L4	Behavioural (sandbox)	SKIPPED

What the source does

Filesystem: declared (8 observation(s))
Network: declared (9 observation(s))
Shell: declared (5 observation(s))
Dependencies: not all pinned
Secrets in source: found

Findings (25)

HIGHPrompt injection · prompt.credential_read · CWE-94, CWE-1427

.claude/skills/1password-sdk/SKILL.md:4

How to load secrets and environment variables from 1Password programmatically

Why it matters. asks the agent to read credentials

HIGHPrompt injection · prompt.credential_read · CWE-94, CWE-1427

.claude/skills/1password-sdk/SKILL.md:6

whenever code needs to read a 1Password secret, resolve an op:// reference,

Why it matters. asks the agent to read credentials

HIGHPrompt injection · prompt.credential_read · CWE-94, CWE-1427

.claude/skills/1password-sdk/SKILL.md:8

(e.g. to import API keys), or read a 1Password Environment — even if the user

Why it matters. asks the agent to read credentials

HIGHPrompt injection · prompt.credential_read · CWE-94, CWE-1427

.claude/skills/1password-sdk/SKILL.md:99

## Read a 1Password Environment (beta)

Why it matters. asks the agent to read credentials

HIGHPrompt injection · prompt.credential_read · CWE-94, CWE-1427

.claude/skills/1password-sdk/SKILL.md:153

| Read a 1Password Environment | ✅ beta | `environments.getVariables`, **beta SDK only** |

Why it matters. asks the agent to read credentials

HIGHPrompt injection · prompt.persistence · CWE-94, CWE-1427

packages/cli/skills/claudish-usage/SKILL.md:788

# Or add to ~/.zshrc or ~/.bashrc

Why it matters. instructs the agent to persist itself in the user's environment

HIGHPrompt injection · prompt.persistence · CWE-94, CWE-1427

skills/claudish-usage/SKILL.md:788

# Or add to ~/.zshrc or ~/.bashrc

Why it matters. instructs the agent to persist itself in the user's environment

MEDIUMInventory / provenance · inv.binary · CWE-1104

packages/cli/src/test-fixtures/devin/GetCliModelConfigs.res.bin

GetCliModelConfigs.res.bin

Why it matters. a compiled or binary member cannot be reviewed from source

Fix. ship source, or explain the binary in the README

MEDIUMInformation disclosure · disclose.log_secret · CWE-209, CWE-532

packages/macos-bridge/src/server.ts:1082

console.log(`CLAUDISH_BRIDGE_TOKEN=${token}`);