langchain-ai/deepagents · error · ValueError

Server '{server_name}' cannot set both 'allowedTools' and 'd

Error message

Server '{server_name}' cannot set both 'allowedTools' and 'disabledTools' — pick one.

What it means

deepagents_code MCP server configs let you filter which tools are exposed either with an allow-list (`allowedTools`) or a block-list (`disabledTools`), but not both at once. `_validate_tool_filter_fields` raises this ValueError during `_validate_server_config` when a single server entry defines both keys, because the combined semantics would be ambiguous. Fix the config so exactly one of the two fields is present per server.

Source

Thrown at libs/code/deepagents_code/mcp_tools.py:988

    (`disabledTools`) — both are almost certainly user errors; omit the field
    instead.

    Args:
        server_name: Name of the server (for error messages).
        server_config: Server configuration dictionary.

    Raises:
        TypeError: If a field is not a list of strings.
        ValueError: If both fields are set, or either field is empty.
    """
    has_allowed = "allowedTools" in server_config
    has_disabled = "disabledTools" in server_config
    if has_allowed and has_disabled:
        error_msg = (
            f"Server '{server_name}' cannot set both 'allowedTools' and"
            " 'disabledTools' — pick one."
        )
        raise ValueError(error_msg)

    for field_name in ("allowedTools", "disabledTools"):
        if field_name not in server_config:
            continue
        value = server_config[field_name]
        if not isinstance(value, list) or not all(
            isinstance(item, str) for item in value
        ):
            error_msg = (
                f"Server '{server_name}' '{field_name}' must be a list of strings"
            )
            raise TypeError(error_msg)
        if not value:
            error_msg = (
                f"Server '{server_name}' '{field_name}' must be non-empty;"
                " omit the field to disable filtering."
            )
            raise ValueError(error_msg)

View on GitHub (pinned to a1af029e6e)

Solutions

Choose one filtering style for the server: keep 'allowedTools' (allow-list) and delete the 'disabledTools' key, or vice versa.
If you need allow-list semantics with exclusions, move the excluded tools out of 'allowedTools' instead of listing them in 'disabledTools'.
If the conflict comes from merged configs, ensure the merge is per-key (last-writer-wins) rather than concatenating keys into the same server dict.

Example fix

// before
{"mcpServers": {"fs": {"command": "mcp-server-fs", "allowedTools": ["read_file"], "disabledTools": ["write_file"]}}}
// after
{"mcpServers": {"fs": {"command": "mcp-server-fs", "allowedTools": ["read_file"]}}}

Defensive patterns

Strategy: validation

Validate before calling

def check_tool_filter(server):
    has_allowed = "allowedTools" in server
    has_disabled = "disabledTools" in server
    if has_allowed and has_disabled:
        raise ValueError("set only one of allowedTools/disabledTools")

Type guard

def has_conflicting_filters(server: dict) -> bool:
    return "allowedTools" in server and "disabledTools" in server

Try / catch

try:
    load_mcp_config(path)
except ValueError as e:
    if "cannot set both" in str(e):
        fix_config_file(path)  # drop one of the two keys
    else:
        raise

Prevention

Pick one filtering convention per project and document it in the config README.
When merging configs, merge per-server dicts wholesale instead of key-by-key.
Add a CI check that validates the MCP config schema before deployment.

When it happens

Trigger: Loading an MCP config (via _load_mcp_config_top_level -> _validate_server_config) where one server dict under 'mcpServers' contains both 'allowedTools' and 'disabledTools' keys, e.g. {"mcpServers": {"fs": {"command": "mcp-server-fs", "allowedTools": ["read_file"], "disabledTools": ["write_file"]}}}.

Common situations: Merging configs from two sources (project .mcp.json plus user config) where one set allowedTools and the other disabledTools; copy-pasting a server entry from docs that used allowedTools and appending a disabledTools exclusion; team members with different filtering conventions editing the same config file.

Related errors

AI-assisted analysis of langchain-ai/deepagents@a1af029e6e (2026-08-29). Data as JSON: /api/errors/b61c861e6b1e1017. Report an issue: GitHub.