A MCP server that enables AI models to invoke AI CLI agents (Gemini CLI, Codex, Claude Code, OpenCode) as tools. Provides parallel execution, automatic retries with exponential backoff, JSON-first response parsing, and structured output through five MCP tools.

Nexus MCP is useful whenever a task benefits from querying multiple AI agents in parallel rather than sequentially:

• Research & summarization — fan out a topic to multiple agents, then synthesize their responses into a single summary with diverse perspectives
• Code review — send different files or review angles (security, correctness, style) to separate agents simultaneously
• Multi-model comparison — prompt the same question to different models and compare outputs side-by-side for quality or consistency
• Bulk content generation — generate multiple test cases, translations, or documentation pages concurrently instead of one at a time
• Second-opinion workflows — get independent answers from separate agents before making a decision, reducing single-model bias

• Parallel execution — batch_prompt fans out tasks with asyncio.gather and a configurable semaphore (default concurrency: 3)
• Automatic retries — exponential backoff with full jitter for transient errors (HTTP 429/503)
• Output handling — JSON-first parsing, brace-depth fallback for noisy stdout, temp-file spillover for outputs exceeding 50 KB
• Execution modes — default (safe, no auto-approve), yolo (full auto-approve)
• CLI detection — auto-detects binary path, version, and JSON output capability at startup
• Session preferences — set defaults for execution mode, model, max retries, output limit, and timeout once per session; subsequent calls inherit them without repeating parameters
• Tool timeouts — configurable safety timeout (default 15 min) cancels long-running tool calls to prevent the server from blocking indefinitely
• Client-visible logging — runner events (retries, output truncation, error recovery) are sent to MCP clients via protocol notifications, not just server stderr
• Elicitation — interactive parameter resolution via MCP elicitation; disambiguates missing CLI, offers model selection, confirms YOLO mode, and prompts for elaboration on vague prompts. Auto-detects client support and skips gracefully when unavailable. Suppression flags prevent repeat prompts within a session
• Extensible — implement build_command + parse_output, register in RunnerFactory

uvx nexus-mcp

uvx installs the package in an ephemeral virtual environment and runs it — no cloning required.

To check the installed version:

uvx nexus-mcp --version

To update to the latest version:

uvx --reinstall nexus-mcp

Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "nexus-mcp": {
      "command": "uvx",
      "args": ["nexus-mcp"],
      "env": {
        "NEXUS_GEMINI_MODEL": "gemini-3-flash-preview",
        "NEXUS_GEMINI_MODELS": "gemini-3.1-pro-preview,gemini-3-flash-preview,gemini-2.5-pro,gemini-2.5-flash,gemini-2.5-flash-lite",
        "NEXUS_CODEX_MODEL": "gpt-5.2",
        "NEXUS_CODEX_MODELS": "gpt-5.4,gpt-5.4-mini,gpt-5.3-codex,gpt-5.2-codex,gpt-5.2,gpt-5.1-codex-max,gpt-5.1-codex-mini",
        "NEXUS_OPENCODE_MODEL": "ollama-cloud/kimi-k2.5",
        "NEXUS_OPENCODE_MODELS": "ollama-cloud/glm-5,ollama-cloud/kimi-k2.5,ollama-cloud/qwen3-coder-next,ollama-cloud/minimax-m2.5,ollama/gemini-3-flash-preview"
      }
    }
  }
}

Cursor (.cursor/mcp.json in your project or ~/.cursor/mcp.json globally):

{
  "mcpServers": {
    "nexus-mcp": {
      "command": "uvx",
      "args": ["nexus-mcp"],
      "env": {
        "NEXUS_GEMINI_MODEL": "gemini-3-flash-preview",
        "NEXUS_GEMINI_MODELS": "gemini-3.1-pro-preview,gemini-3-flash-preview,gemini-2.5-pro,gemini-2.5-flash,gemini-2.5-flash-lite",
        "NEXUS_CODEX_MODEL": "gpt-5.2",
        "NEXUS_CODEX_MODELS": "gpt-5.4,gpt-5.4-mini,gpt-5.3-codex,gpt-5.2-codex,gpt-5.2,gpt-5.1-codex-max,gpt-5.1-codex-mini",
        "NEXUS_OPENCODE_MODEL": "ollama-cloud/kimi-k2.5",
        "NEXUS_OPENCODE_MODELS": "ollama-cloud/glm-5,ollama-cloud/kimi-k2.5,ollama-cloud/qwen3-coder-next,ollama-cloud/minimax-m2.5,ollama/gemini-3-flash-preview"
      }
    }
  }
}

Claude Code (CLI):

claude mcp add nexus-mcp \
  -e NEXUS_GEMINI_MODEL=gemini-3-flash-preview \
  -e NEXUS_GEMINI_MODELS=gemini-3.1-pro-preview,gemini-3-flash-preview,gemini-2.5-pro,gemini-2.5-flash,gemini-2.5-flash-lite \
  -e NEXUS_CODEX_MODEL=gpt-5.2 \
  -e NEXUS_CODEX_MODELS=gpt-5.4,gpt-5.4-mini,gpt-5.3-codex,gpt-5.2-codex,gpt-5.2,gpt-5.1-codex-max,gpt-5.1-codex-mini \
  -e NEXUS_OPENCODE_MODEL=ollama-cloud/kimi-k2.5 \
  -e NEXUS_OPENCODE_MODELS=ollama-cloud/glm-5,ollama-cloud/kimi-k2.5,ollama-cloud/qwen3-coder-next,ollama-cloud/minimax-m2.5,ollama/gemini-3-flash-preview \
  -- uvx nexus-mcp

Generic stdio config (any MCP-compatible client):

{
  "command": "uvx",
  "args": ["nexus-mcp"],
  "transport": "stdio",
  "env": {
    "NEXUS_GEMINI_MODEL": "gemini-3-flash-preview",
    "NEXUS_CODEX_MODEL": "gpt-5.2",
    "NEXUS_OPENCODE_MODEL": "ollama-cloud/kimi-k2.5"
  }
}

All env keys are optional — see Configuration for the full list.

Prerequisites:

• Python 3.13+ (download)
• uv dependency manager (install guide)

  curl -LsSf https://astral.sh/uv/install.sh | sh

Optional (for integration tests):

• Gemini CLI v0.6.0+ — npm install -g @google/gemini-cli
• Codex — check with codex --version
• Claude Code — check with claude --version
• OpenCode — check with opencode --version

Note: Integration tests are optional. Unit tests run without CLI dependencies via subprocess mocking.

# 1. Clone the repository
git clone <repository-url>
cd nexus-mcp

# 2. Install dependencies
uv sync

# 3. Install pre-commit hooks (runs linting/formatting on commit)
uv run pre-commit install

# 4. Verify installation
uv run pytest                    # Run tests
uv run mypy src/nexus_mcp        # Type checking
uv run ruff check .              # Linting

# 5. Run the MCP server
uv run python -m nexus_mcp

Once nexus-mcp is configured in your MCP client, your AI assistant automatically sees its tools. The reliable trigger is explicitly asking for output from an external AI agent (e.g. Gemini, Codex, Claude Code, OpenCode). Generic "do this in parallel" prompts may be handled by the host AI's own capabilities instead. The cli parameter is optional — if omitted and the client supports MCP elicitation, the server will ask which runner to use. The server provides runner metadata (names, models, availability, execution modes) in its connection instructions — no discovery call needed. The cli parameter includes a JSON schema enum listing valid runner names.

You say to your AI assistant:

"Get perspectives from Gemini, Codex, and OpenCode on transformer architectures — summary, limitations, and applications."

Your AI assistant calls batch_prompt with the discovered runners:

{
  "tasks": [
    { "cli": "gemini", "prompt": "Summarize the key findings of the Attention Is All You Need paper", "label": "gemini-summary" },
    { "cli": "codex", "prompt": "What are the main limitations of transformer architectures?", "label": "codex-limitations" },
    { "cli": "opencode", "prompt": "List 3 real-world applications of transformers beyond NLP", "label": "opencode-applications" }
  ]
}

You say to your AI assistant:

"Have Gemini, Codex, and OpenCode each review this diff in parallel — I want three independent perspectives."

Your AI assistant calls batch_prompt:

{
  "tasks": [
    { "cli": "gemini", "prompt": "Review this diff for security vulnerabilities and logic errors:\n\n<paste diff>", "label": "gemini-review" },
    { "cli": "codex", "prompt": "Review this diff for correctness and edge cases:\n\n<paste diff>", "label": "codex-review" },
    { "cli": "opencode", "prompt": "Review this diff for style and maintainability:\n\n<paste diff>", "label": "opencode-review" }
  ]
}

You say to your AI assistant:

"Ask Gemini Flash to explain the difference between TCP and UDP in simple terms."

Your AI assistant calls prompt:

{
  "cli": "gemini",
  "prompt": "Explain the difference between TCP and UDP in simple terms",
  "model": "gemini-3-flash-preview"
}

Or target Codex:

{
  "cli": "codex",
  "prompt": "Explain the difference between TCP and UDP in simple terms",
  "model": "gpt-5.2"
}

Or OpenCode:

{
  "cli": "opencode",
  "prompt": "Explain the difference between TCP and UDP in simple terms",
  "model": "ollama-cloud/kimi-k2.5"
}

You say to your AI assistant:

"Explain the CAP theorem using one of the available agents."

Your AI assistant calls prompt without specifying cli:

{
  "prompt": "Explain the CAP theorem in simple terms"
}

If the MCP client supports elicitation, the server asks which runner to use. If elicitation is unavailable, the server returns an error asking for cli to be specified. To skip elicitation for a specific call, pass "elicit": false.

You say to your AI assistant:

"For the rest of this session, use YOLO mode with Gemini Flash — I don't want to repeat those settings on every call."

Your AI assistant calls set_preferences once:

{
  "execution_mode": "yolo",
  "model": "gemini-3-flash-preview",
  "max_retries": 5
}

Response:

Preferences set: {"execution_mode": "yolo", "model": "gemini-3-flash-preview", "max_retries": 5, "output_limit": null, "timeout": null, "retry_base_delay": null, "retry_max_delay": null, "elicit": null, "confirm_yolo": null, "confirm_vague_prompt": null, "confirm_high_retries": null, "confirm_large_batch": null}

Subsequent prompt and batch_prompt calls omit those fields — they inherit from the session:

{
  "cli": "gemini",
  "prompt": "Summarize the latest developments in Rust's async ecosystem"
}

The fallback chain is: explicit parameter → session preference → per-runner env → global env → hardcoded default. To override for one call, pass the parameter directly — it takes precedence without changing the session. To clear a single preference, use set_preferences with the corresponding clear_* flag (e.g. clear_model: true). Other preferences are preserved.

All prompt tools run as background tasks — they return a task ID immediately so the client can poll for results, preventing MCP timeouts for long operations (e.g. YOLO mode: 2–5 minutes).

Task object fields:

Note: elicit is a batch-level parameter, not a per-task field. When enabled, the server runs a single upfront elicitation pass across all tasks (e.g., "3 of 5 tasks use YOLO mode — confirm?") rather than prompting per-task.

No parameters. Returns a dict with all preference fields (null when unset):

No parameters. Resets all session preferences to null, including elicitation suppression flags (re-enables all confirmation prompts).

Pattern: NEXUS_{AGENT}_{KEY} (agent name uppercased). Per-runner values override global values.

Valid {AGENT} values: CLAUDE, CODEX, GEMINI, OPENCODE

Invalid per-runner values are silently ignored (the global or hardcoded default is used instead).

This project follows Test-Driven Development (TDD) with strict Red→Green→Refactor cycles.

# Run all tests
uv run pytest

# Run with coverage report
uv run pytest --cov=nexus_mcp --cov-report=term-missing

# Run specific test types
uv run pytest -m integration           # Integration tests (requires CLIs)
uv run pytest -m "not integration"     # Unit tests only
uv run pytest -m "not slow"            # Skip slow tests

# Run specific test file
uv run pytest tests/unit/runners/test_gemini.py

Test markers:

• @pytest.mark.integration — requires real CLI installations
• @pytest.mark.slow — tests taking >1 second

All quality checks run automatically via pre-commit hooks. Run manually:

# Lint and format
uv run ruff check .              # Check for issues
uv run ruff check --fix .        # Auto-fix issues
uv run ruff format .             # Format code

# Type checking (strict mode)
uv run mypy src/nexus_mcp

# Run all pre-commit hooks manually
uv run pre-commit run --all-files

uv add <package>              # Production dependency
uv add --dev <package>        # Development dependency
uv sync                       # Sync environment after changes

• Ruff: line length 100, 17 rule sets (E/F/I/W + UP/FA/B/C4/SIM/RET/ICN/TID/TC/ISC/PTH/TD/NPY) — pyproject.toml → [tool.ruff]
• Mypy: strict mode, all type annotations required — pyproject.toml → [tool.mypy]
• Pytest: asyncio_mode = "auto", no @pytest.mark.asyncio needed — pyproject.toml → [tool.pytest.ini_options]
• Pre-commit: ruff-check, ruff-format, mypy, trailing-whitespace, end-of-file-fixer — .pre-commit-config.yaml

• type keyword for type aliases: type AgentName = str
• Union syntax: str | None (not Optional[str])
• match statements for complex conditionals
• NO from __future__ import annotations

nexus-mcp/
├── src/nexus_mcp/
│   ├── __main__.py         # Entry point
│   ├── server.py           # FastMCP server + tools
│   ├── types.py            # Pydantic models
│   ├── exceptions.py       # Exception hierarchy
│   ├── config.py           # Environment variable config
│   ├── elicitation.py      # ElicitationGuard — interactive parameter resolution
│   ├── process.py          # Subprocess wrapper
│   ├── parser.py           # JSON→text fallback parsing
│   ├── cli_detector.py     # CLI binary detection + version checks
│   └── runners/
│       ├── base.py         # Protocol + ABC
│       ├── factory.py      # RunnerFactory
│       ├── claude.py       # ClaudeRunner
│       ├── codex.py        # CodexRunner
│       ├── gemini.py       # GeminiRunner
│       └── opencode.py     # OpenCodeRunner
├── tests/
│   ├── unit/               # Fast, mocked tests
│   ├── e2e/                # End-to-end MCP protocol tests
│   ├── integration/        # Real CLI tests
│   └── fixtures.py         # Shared test utilities
├── .github/
│   └── workflows/          # CI, security, dependabot
├── pyproject.toml          # Dependencies + tool config
└── .pre-commit-config.yaml # Git hooks configuration

MIT