Atlas / MCP servers / sluisr / 🐋 DeepSeek CLI

🐋 DeepSeek CLIBLOCK

Name: 🐋 DeepSeek CLI
Author: sluisr

mcp/sluisr/deepseek-cli

MCP server by sluisr · Audited 2026-10-12

⚑ Report / re-scan Source ↗

DeepSeek CLI — unofficial adaptation of Gemini CLI by sluisr. Full agentic AI in your terminal using DeepSeek API.

Verdict

BLOCK

Grade

Trust score

21 /100

Exposed tools

200 238r · 23w · 2d

Transport

stdio · streamable-http

License

Apache-2.0

Stars

Overview

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

Unofficial fork of Google Gemini CLI adapted to use the DeepSeek API. Original work Copyright 2025 Google LLC — Adaptations Copyright 2026 sluisr — Apache 2.0 License.

[](LICENSE) [](https://www.npmjs.com/package/@sluisr/deepseek-cli)

DeepSeek CLI is an open-source AI agent that brings the power of DeepSeek directly into your terminal. High-performance autonomous coding, tool execution, and context management by sluisr.

🚀 Why DeepSeek CLI?

💰 Pay-per-use & Cost-Efficient: Pay only for tokens used via DeepSeek API with up to 90% KV Cache discount.
🧠 DeepSeek V4 Models: Full support for deepseek-v4-flash and deepseek-v4-pro (Reasoning / Thinking CoT mode).
⚡ Fast Code Patching: Atomic unified diff patching (apply_patch) for instant, token-efficient code modifications.
🛠️ Built-in Developer Tools: File system operations, native server-side web search, shell execution, memory persistence, and FIM autocompletion.
🔑 Silent Sudo & System Auth: Native non-blocking zero-lag AskPass integration for sudo, ssh, and git.
🔌 MCP Extensible: Full Model Context Protocol (MCP) server support.
🔒 Isolated Configuration: Operates under ~/.deepseek/ without polluting system or official CLI configurations.

📦 Installation

Quick Install (NPM Global)

npm install -g @sluisr/deepseek-cli

Then run:

deepseek

Update to Latest Version

npm install -g @sluisr/deepseek-cli@latest

🔐 Authentication

Get your API key at platform.deepseek.com/api_keys and export it:

export DEEPSEEK_API_KEY="your-api-key-here"
deepseek

Or enter it interactively when prompted on first launch.

Optional custom base URL or pr

Read from source at commit 8c6eb2a380c4OBSERVED · 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 ingestion-service --env DEEPSEEK_SUDO_PASSWORD=${DEEPSEEK_SUDO_PASSWORD} --env FORCE_GENERIC_KEYBINDING_HINTS=${FORCE_GENERIC_KEYBINDING_HINTS} --env GEMINI_API_KEY=${GEMINI_API_KEY} --env GEMINI_CLI_CREDENTIALS_PATH=${GEMINI_CLI_CREDENTIALS_PATH} -- npx -y [email protected]

claude-desktop

{
  "mcpServers": {
    "ingestion-service": {
      "command": "npx",
      "args": [
        "-y",
        "[email protected]"
      ],
      "env": {
        "DEEPSEEK_SUDO_PASSWORD": "${DEEPSEEK_SUDO_PASSWORD}",
        "FORCE_GENERIC_KEYBINDING_HINTS": "${FORCE_GENERIC_KEYBINDING_HINTS}",
        "GEMINI_API_KEY": "${GEMINI_API_KEY}",
        "GEMINI_CLI_CREDENTIALS_PATH": "${GEMINI_CLI_CREDENTIALS_PATH}"
      }
    }
  }
}

Exposed tools (200)

238 read · 23 write · 2 destructive. Blast radius: 2 tools can delete or overwrite — an agent that can be talked into calling a tool can be talked into calling this one.

Tool	Risk	Description
`BrowserLikeAgent`	read	An agent using instance tools.
`CodebaseInvestigator`	read	Investigates codebase
`Default`	read	Prompts for approval
`DisabledAgent`	read	I am disabled
`Edit`	write	packages/.../InputPrompt.tsx: return kittyProtocolSupporte... => return kittyProtocolSupporte...
`EmptyDescAgent`	read
`Expansion`	read	Command expansion needs shell access
`FailAgent`	read	An agent that fails to load
`FailingRemoteAgent`	read	A remote agent
`Free`	read	free tier
`MixedAgent`	read	Uses both patterns.
`MockAgent`	read	A mock agent for testing.
`OtherAgent`	read	Another agent
`OverwrittenAgent`	read	A remote agent
`Plan`	read	Read-only mode
`ProjectAgent`	read	Project Agent Desc
`R1`	read	D1
`RemoteAgent`	read	A remote agent
`RemoteAgentAccumulationTest`	read	User-provided description
`RemoteAgentBadAuth`	read	A remote agent
`RemoteAgentWithAuth`	read	A remote agent
`RemoteAgentWithDescription`	read	User-provided description
`RemoteAgentWithEmptyAgentDescription`	read	User-provided description
`RemoteAgentWithSkillsOnly`	read	User-provided description
`RemotePolicyAgent`	read	A remote agent
`SecuredAgent`	read	A secured remote agent
`Shell`	read	user-command
`ShellTool`	read	A shell tool
`Skill1`	read	Desc1
`Skill2`	read	Desc2
`SkillFallbackAgent`	read	User description
`Test`	read	Test desc
`TestAgent`	read	An agent for testing.
`TestLocalAgent`	read	A local test agent.
`TestProtocolAgent`	read	A test agent for protocol tests.
`TestRemoteAgent`	read	A remote test agent.
`Username`	read	Your public username
`Value`	read	Val
`YOLO`	read	Auto-approves all tools
`__leading`	read	test
`a__b__c`	read	three parts - not valid MCP
`about`	read	About the CLI
`accessibility-agent`	read	An expert in web accessibility and ARIA roles.
`add`	write	Add two numbers
`agent_1`	read	Test agent 1
`agent_2`	read	Test agent 2
`agents`	read	Manage agents
`alpha`	read	Alpha command
`analytics-agent`	read	An expert in tracking, analytics, and metrics.
`another-command`	read	Another test command
`another-tool`	read	Another tool
`arg1`	read	Argument 1
`ask_user`	read	ask user
`audit`	read	Triage issues labeled as status/need-triage
`auth`	read	Manage authentication
`bravo`	read	Bravo command
`browser_agent`	read	mock definition
`bug`	write	Submit a bug report
`bug-memory`	read	Capture a V8 heap snapshot to disk to attach to a bug report
`builtin`	read	desc2
`builtin-skill`	read	A built-in skill
`catchAllTool`	read	Catches all errors
`chat`	read	Browse auto-saved conversations and manage chat checkpoints
`checkSystemStatus`	read	Checks the current system status
`child`	read	a child command
`child1`	read	First child.
`clear`	destructive	Clear the screen
`click`	read	Click element
`click_at`	read	Click at coordinates
`closeDiff`	read
`close_page`	read	Close page
`codebase_investigator`	read	Investigating codebase
`commands`	read	Manage custom slash commands. Usage: /commands [list\|reload]
`commit`	write	Commit changes
`common-agent`	read	User version
`complex-agent`	read	A complex markdown agent
`config`	read	Configure an agent
`confirm`	read	a command that needs confirmation
`confirming_tool`	read	Confirming tool description
`context-check-command`	read	checks context
`copy`	read	Copy the last result or code snippet to clipboard
`corgi`	read	Toggles corgi mode
`css-agent`	read	An expert in CSS, styling, and UI design.
`custom-script`	write	Run custom script
`cyclic-command`	read
`database-agent`	read	An expert in database schemas, SQL, and creating database migrations.
`debug`	read	Export the most recent API request as a JSON payload
`dedup`	read	Triage issues labeled as status/possible-duplicate
`delete`	destructive	Delete a conversation checkpoint. Usage: /resume delete <tag>
`devops-agent`	read	An expert in CI/CD, Docker, and deployment scripts.
`disable`	write	Disable an enabled agent
`disabled-builtin`	read	disabled builtin
`docs`	read	Open full Gemini CLI documentation in your browser
`docs-agent`	read	An agent with expertise in updating documentation.
`edit`	write	edit file
`edit-tool`	write	desc
`editor`	write	Set external editor preference
`enable`	write	Enable a disabled agent
`error-tool`	read	This tool failed
`exit_plan_mode`	read	exit plan mode
`explore`	read	Open extensions page in your browser
`export-session`	read	Export the current session to a JSON file
`extensions`	read	Manage extensions.
`failTool`	read	Mock description
`failVisible`	read	Fails with a visible error if input is
`failing-tool`	read	A tool that fails
`fetch_posts`	read
`filecmd`	read	A command from a file
`fill`	read	Fill form field
`foo`	read	example resource
`free`	read	Free tier
`gemma`	read	Check local Gemma model routing status
`get_context`	read	Get information about the current session context.
`get_ticket`	read	Get a ticket
`help`	read	Show help for this prompt
`hidden`	read	A hidden command
`hidden-child`	read	A hidden child command
`hooks`	read	Manage hooks
`hover`	read	Hover element
`i18n-agent`	read	An expert in internationalization and translations.
`ide`	read	IDE command
`inbox`	write	Review skills extracted from past sessions and move them to global or project skills
`inferred-remote`	read	Inferred
`init`	read	Initializes the server.
`install`	write	Install an extension from a git repo or local path
`json-remote`	read	A JSON-based remote agent
`legacy-tool`	read	desc
`link`	read	Link an extension from a local path
`list`	read	List available local and remote agents
`list_pages`	read	list all pages
`locked-skill`	read	A skill in a locked dir
`ls`	read	list directory
`manage`	read	Opens a dialog to configure the model
`mcp`	read	MCP command
`mcp-agent`	read	An agent with MCP servers
`mcp_myserver_search`	read	Search via MCP
`mcpcmd`	read	A command from mcp
`memory`	read	Commands for interacting with memory
`memory-leak`	read	Debug memory leaks
`mobile-agent`	read	An expert in React Native and mobile app development.
`mock-agent`	read	Mock Agent Description
`model`	read	Manage model configuration
`multiAnnotationTool`	read	A tool with multiple annotations
`navigate_page`	read	Navigate to URL
`new_page`	read	Open a new page/tab
`no-card-agent`	read	Missing card info
`noaction`	read	unhandled type
`non-stream-test`	read	A test non-streaming command
`not-top-level`	read	Not a top level command
`notakes`	read	Subcommand that does not take arguments
`nullableTool`	read	Tool with nullable array
`oauth-test-agent`	read	An agent to test OAuth MCP with full fields
`oncall`	read	Oncall related commands
`one`	read	first
`openDiff`	read
`other_tool`	read
`paid`	read	Paid tier
`parent`	read	A parent command
`pending-tool`	read	This tool is pending
`permissions`	read	Permissions command
`plan`	read	Plan command
`policies`	read	Manage policies
`policy-test-tool`	read	tests policy
`press_key`	read	Press key
`privacy`	read	Display the privacy notice
`profile`	read	Profile command
`project-agent`	read	desc
`project-gemini`	read	desc
`project-only`	read	Project only
`prompt-1`	read	A test prompt
`question-mark`	read	Alternative name for help
`readOnlyTool`	read	A read-only tool
`read_file`	read	Read a file
`regular`	read	desc1
`regular-skill`	read	regular
`remote-agent`	read	A remote agent
`remote_agent`	read	Remote agent description.
`remote_agent_call`	read	Remote agent call
`replace`	write	edit file
`resource`	read	Test Resource
`restore`	read	Restores the server.
`resume`	read	Browse auto-saved conversations and manage chat checkpoints
`review`	read	Review code
`review-frontend`	read	Review frontend code
`rewind`	write	Jump back to a specific message and restart the conversation
`run_shell_command`	write	Run command
`s1`	read	d1
`s2`	read	d2
`s3`	read	d3
`same-name`	read	ext-desc
`same-skill`	read	gemini-desc
`save`	write	Save the current conversation as a checkpoint. Usage: /resume save <tag>
`schema`	read	List configured MCP servers and tools with descriptions and schemas
`security-agent`	read	An expert in security audits and vulnerability patches.
`select_page`	read	Select page
`session`	read	Show session-specific usage statistics
`set`	write	Set the model to use. Usage: /model set <model-name> [--persist]
`settings`	write	View and edit Gemini CLI settings
`setup-github`	write	Set up GitHub Actions
`share`	read	Share the current conversation to a markdown or json file. Usage: /resume share <file>

Trust audit

BLOCKgrade F · trust 21/100 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	FAIL
L2	Instruction surface (what it tells the agent)	FAIL
L3	Class-specific surface	WARN
L4	Behavioural (sandbox)	SKIPPED

What the source does

Filesystem: declared (8 observation(s))
Network: declared (13 observation(s))
Shell: declared (5 observation(s))
Dependencies: not all pinned
Secrets in source: found

Findings (25)

HIGHCode injection · code.eval_exec · CWE-78, CWE-94, CWE-95

packages/cli/src/acp/commands/about.ts:32

sandboxEnv = `sandbox-exec (${

Why it matters. evaluates text as code

Fix. remove; use a parser or a dispatch table

HIGHCode injection · code.eval_exec · CWE-78, CWE-94, CWE-95

packages/cli/src/ui/commands/aboutCommand.ts:33

sandboxEnv = `sandbox-exec (${

Why it matters. evaluates text as code

Fix. remove; use a parser or a dispatch table

HIGHCode injection · code.eval_exec · CWE-78, CWE-94, CWE-95

packages/cli/src/ui/commands/bugCommand.ts:46

sandboxEnv = `sandbox-exec (${

Why it matters. evaluates text as code

Fix. remove; use a parser or a dispatch table

HIGHCode injection · code.eval_exec · CWE-78, CWE-94, CWE-95

packages/sdk/src/types.ts:183

exec(cmd: string, options?: AgentShellOptions): Promise<AgentShellResult>;

Why it matters. evaluates text as code

Fix. remove; use a parser or a dispatch table

HIGHNetwork egress · net.env_exfil · CWE-200, CWE-319

packages/cli/src/utils/sandbox.ts:170

{ ...process.env } ... curl

Why it matters. reads secrets in the same file that sends data out

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

docs/changelogs/index.md:914

- **Experimental todos:** Track and display progress on complex tasks with a

Why it matters. invisible characters in instruction text

Fix. strip non-printing characters

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

packages/vscode-ide-companion/NOTICES.txt:2217

# MIT License

Why it matters. invisible characters in instruction text

Fix. strip non-printing characters

MEDIUMInventory / provenance · inv.binary · CWE-1104

packages/core/src/utils/__fixtures__/dummy.wasm

dummy.wasm

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

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

MEDIUMNetwork egress · net.raw_ip · CWE-200, CWE-319

integration-tests/browser-agent-localhost.dynamic.responses:1

{"method":"generateContentStream","response":[{"candidates":[{"content":{"parts":[{"text":"I'll check the dynamic content page on the localhost server."},{"functionCall":{"name":"invoke_agent","args":

MEDIUMNetwork egress · net.raw_ip · CWE-200, CWE-319

integration-tests/browser-agent-localhost.dynamic.responses:2

{"method":"generateContentStream","response":[{"candidates":[{"content":{"parts":[{"functionCall":{"name":"navigate_page","args":{"url":"http://127.0.0.1:18923/dynamic.html"}}}],"role":"model"},"finis

MEDIUMNetwork egress · net.raw_ip · CWE-200, CWE-319

integration-tests/browser-agent-localhost.form.responses:1

{"method":"generateContentStream","response":[{"candidates":[{"content":{"parts":[{"text":"I'll fill out the contact form on the localhost server."},{"functionCall":{"name":"invoke_agent","args":{"age

MEDIUMNetwork egress · net.raw_ip · CWE-200, CWE-319

integration-tests/browser-agent-localhost.form.responses:2

{"method":"generateContentStream","response":[{"candidates":[{"content":{"parts":[{"functionCall":{"name":"navigate_page","args":{"url":"http://127.0.0.1:18923/form.html"}}}],"role":"model"},"finishRe

MEDIUMNetwork egress · net.raw_ip · CWE-200, CWE-319

integration-tests/browser-agent-localhost.multistep.responses:1

{"method":"generateContentStream","response":[{"candidates":[{"content":{"parts":[{"text":"I'll go through the multi-step flow on the localhost server."},{"functionCall":{"name":"invoke_agent","args":

MEDIUMObfuscation / stealth · obf.anti_debug · CWE-506, CWE-94

packages/core/src/sandbox/linux/LinuxSandboxManager.ts:84

{ code: 0x15, jt: 0, jf: 1, k: SYS_ptrace }, // If ptrace, jump to ERRNO