Skip to main content
These commands let you explore and exercise the mirrored command and tool inventories that Claw Code builds from the archived TypeScript snapshots.

commands

Lists mirrored command entries from the archived snapshot.
python3 -m src.main commands [--limit N] [--query Q] [--no-plugin-commands] [--no-skill-commands]

Arguments

--limit
integer
default:"20"
Maximum number of entries to display.
--query
string
Search string. When provided, the command calls render_command_index() which performs a fuzzy match and returns a ranked Markdown index instead of the plain list.
--no-plugin-commands
boolean
default:"false"
Exclude plugin-sourced commands from the listing.
--no-skill-commands
boolean
default:"false"
Exclude skill-sourced commands from the listing.

Output format

Without --query:
Command entries: 42

- /compact — commands/compact.ts
- /config — commands/config.ts
- /doctor — commands/doctor.ts
- /help — commands/help.ts
- /init — commands/init.ts
With --query:
## Command index — query: "config"

- /config — commands/config.ts
- /init — commands/init.ts

tools

Lists mirrored tool entries from the archived snapshot.
python3 -m src.main tools [--limit N] [--query Q] [--simple-mode] [--no-mcp] [--deny-tool NAME] [--deny-prefix PREFIX]

Arguments

--limit
integer
default:"20"
Maximum number of entries to display.
--query
string
Search string. When provided, calls render_tool_index() for a ranked Markdown result.
--simple-mode
boolean
default:"false"
Load the tool inventory in simple mode (reduced tool set, no advanced tool variants).
--no-mcp
boolean
default:"false"
Exclude MCP-sourced tools from the listing.
--deny-tool
string
Deny a specific tool by exact name. Can be repeated to deny multiple tools.
python3 -m src.main tools --deny-tool Bash --deny-tool Write
--deny-prefix
string
Deny all tools whose names start with a given prefix. Can be repeated.
python3 -m src.main tools --deny-prefix mcp_

Output format

Without --query:
Tool entries: 18

- Bash — tools/bash.ts
- Read — tools/read.ts
- Write — tools/write.ts
- Edit — tools/edit.ts
- Glob — tools/glob.ts

show-command

Shows one mirrored command entry by exact name.
python3 -m src.main show-command <name>

Arguments

name
string
required
Exact name of the command to look up (e.g. /config).

Output format

/config
commands/config.ts
Manages local and global Claude configuration settings
Three lines: name, source hint, responsibility description. Returns exit code 1 if the name is not found.

show-tool

Shows one mirrored tool entry by exact name.
python3 -m src.main show-tool <name>

Arguments

name
string
required
Exact name of the tool to look up (e.g. Bash).

Output format

Bash
tools/bash.ts
Executes shell commands in a sandboxed environment
Three lines: name, source hint, responsibility description. Returns exit code 1 if the name is not found.

exec-command

Executes a mirrored command shim by exact name.
python3 -m src.main exec-command <name> <prompt>

Arguments

name
string
required
Exact name of the command to execute.
prompt
string
required
The prompt string to pass to the command shim’s execute() method.

Output format

Prints the message field from the ExecutionResult. Returns exit code 0 when the command reports handled=True, or 1 otherwise.
python3 -m src.main exec-command /config "show current model"
[/config] executed with prompt: show current model

exec-tool

Executes a mirrored tool shim by exact name.
python3 -m src.main exec-tool <name> <payload>

Arguments

name
string
required
Exact name of the tool to execute.
payload
string
required
A string payload passed to the tool shim’s execute() method. For tools that expect structured input, pass a JSON string.

Output format

Prints the message field from the ExecutionResult. Returns exit code 0 when handled=True, or 1 otherwise.
python3 -m src.main exec-tool Read '{"file_path": "src/main.py"}'
[Read] executed with payload: {"file_path": "src/main.py"}

route

Routes a prompt across the full command and tool inventories and returns ranked matches.
python3 -m src.main route <prompt> [--limit N]

Arguments

prompt
string
required
Free-text prompt to route. The router tokenises the prompt, scores each command and tool by counting token matches against the entry’s name, source hint, and responsibility fields, then returns the top matches.
--limit
integer
default:"5"
Maximum number of matches to return. The router always tries to include at least one command match and one tool match before filling remaining slots with the highest-scoring leftovers.

Output format

Each line is tab-separated:
<kind>\t<name>\t<score>\t<source_hint>
python3 -m src.main route "read a file" --limit 3
tool	Read	2	tools/read.ts
command	/help	1	commands/help.ts
tool	Glob	1	tools/glob.ts
If no matches are found the output is:
No mirrored command/tool matches found.
route is the same scoring logic used internally by bootstrap and turn-loop. Running it directly lets you inspect and tune the routing behaviour without triggering a full runtime session.

Build docs developers (and LLMs) love