Model Context Protocol moderate risk

Serena

Serena is an open-source coding agent toolkit that provides LLMs with IDE-like semantic code retrieval and editing capabilities, enhancing their ability to work directly on codebases.

Connections & Capabilities

Connects To

GitHub

Capabilities

readwriteexec

Quickstart

Install

uvx --from git+https://github.com/oraios/serena serena start-mcp-server --help

Config

{
    "mcpServers": {
        "serena": {
            "command": "/abs/path/to/uv",
            "args": ["run", "--directory", "/abs/path/to/serena", "serena-mcp-server"]
        }
    }
}

Exposed MCP Tools (3)

safe

find_symbol

Finds a code symbol (e.g., function, class) within the codebase.

Read-only operation; no code modification.

safe

find_referencing_symbols

Finds all symbols that reference a given symbol.

Read-only operation; no code modification.

moderate

insert_after_symbol

Inserts code after a specified symbol.

Modifies code by inserting new content.

Safety Assessment

Serena's safety depends on the configuration and the LLM using it. It's relatively safe for read-only operations and code analysis. However, code modification capabilities introduce risks, especially if the LLM is not properly controlled or the underlying language servers have vulnerabilities.

Relies on language server protocol (LSP) or JetBrains plugin for code analysis.
Provides semantic code retrieval rather than raw file access.
MCP server architecture allows for controlled tool execution.
Open-source and auditable code base.

Can perform code modifications, potentially introducing errors.
Depends on the security of underlying language servers or JetBrains IDE.
Improperly configured LLM could lead to unintended code changes.
Requires careful configuration to prevent excessive or incorrect modifications.
Access to the codebase is granted to the LLM, posing a risk if the LLM is compromised.