Skip to main content
One command points a tool at a model and starts it. There is nothing to configure by hand:
That runs a model we host, so sign in first with wally account login. Everywhere else, -m takes either kind of id: Wally looks for the model on this machine first and falls back to the hosted catalog. Without -m the tool starts the way you already have it configured, and Wally wires nothing.

​
Code in the terminal

Claude Code

Anthropic’s terminal agent, pointed at any model you can run.

Codex

OpenAI’s terminal coding agent, on your model.

Command Code

A terminal coding agent that learns your coding taste.

OpenCode

An open-source coding TUI, local or hosted.

Hermes Agent

Nous Research’s agent, with its own tools and memory.

DeepSeek Harness

The dsh harness and its browser UI.

Prime Agent

Prime Intellect’s agent, local or hosted.

​
Security tools

These three scan or pentest systems you own. They are hosted-only. Pass the model with -m. The examples below use deepseek-v4.1-flash-1017-uncensored.
That model is gated. If your account does not already have access, the command will not work. Do not pass it unless it is on your account. Use these tools only against repos and apps you own or have written permission to test. The acceptable use policy applies to every request.

DeepSec

Scan your own repos with Vercel’s AI security reviewer.

VVAH

Scan your own repos with Visa’s vulnerability harness.

Shannon

Test your own apps with Shannon, an AI pentester.

​
Connect an assistant

Claude Desktop

The desktop app’s chat and Cowork tabs, on your model.

OpenClaw

A personal agent that runs its own runtime locally.

​
How the wiring works

Every integration gets the same thing: one OpenAI-compatible base URL. The tool never learns whether the model behind it is on this machine or on the console — for a local model the URL is a server Wally starts and stops around the session, and for a hosted one it is the console’s own endpoint. That is why none of these tools needs a Wally plugin. Two tools speak Anthropic’s Messages API rather than OpenAI’s, so a translator sits between them and the model. It carries tool definitions out, tool calls back, and tool results out again, which is what lets the agent actually run the tools it was given. None of it outlives the session. A config Wally writes is scoped to the child process, and a desktop app it reconfigures is put back when the app quits.

​
Images

New In the hosted catalog, a model takes pictures when GET /v1/models lists image in input_modalities — today deepseek-v4.1-flash, deepseek-v4.1-flash-fast, and kimi-k3. Text-only models reject them, and a model on this machine does not either unless that build supports vision. Each harness below says whether it passes them.

​
Flags

​
string
The model to run against. Wally resolves it on this machine first and falls back to the hosted catalog, so a console model id works here on its own. Omit it and the tool starts on whatever it is already configured for.
​
flag
wally opencode only. Go straight to the signed-in hosted endpoint, and never route a local model.
​
integer
wally codex and wally chatgpt only. Tell the tool the model holds this many tokens, so it compacts sooner.
​
flag
wally claude-code, wally claude-desktop and wally codex only. Hold the endpoint open and print it instead of launching anything. This is how you wire up a tool Wally has not been taught about.
​
flag
wally claude-desktop only. Put Claude Desktop back the way it was and start nothing. A normal run already undoes its own configuration when the app quits, so this is for the run that was interrupted first.
Anything after the flags is passed through to the tool untouched. Set the model a harness uses when you pass no -m:

​
Claude Code

Claude Code is Anthropic’s terminal agent. You need the claude binary on your PATH. Wally sets the base URL and token for that process only, and a translator turns its messages into the model’s API. Your normal claude sessions stay untouched. New Paste a picture, or let Read open one. On a model that takes no images, a pasted picture is refused and the next message still works.

​
Codex

Codex is OpenAI’s terminal coding agent. You need the codex binary on your PATH. Anything after -- goes to Codex untouched. Add --context-window with the model’s token count so Codex compacts sooner. To open ChatGPT’s Codex app on a model instead, run wally chatgpt -m glm-5.3-flash.

​
Command Code

Command Code is a terminal coding agent that learns your coding taste. Install it with npm i -g command-code if it is missing. Wally starts it on the model you named, and anything after the flags goes to the tool untouched.

​
OpenCode

OpenCode is an open-source coding TUI. Install it with npm i -g opencode-ai if it is missing. Wally passes the provider through OPENCODE_CONFIG_CONTENT, not your own opencode config. --cloud stays on the hosted model. New Pass -f photo.png, or let the agent open the image. Only a model that takes images will see it.

​
Hermes Agent

Hermes is Nous Research’s agent, with its own tools and memory. Wally starts the TUI and points it at the model. It cannot set Hermes’s context window, so if you want the full window, add model.context_length to your own ~/.hermes/config.yaml. Wally prints the number when it starts. Hermes does not pass images.

​
DeepSeek Harness

The DeepSeek harness is the dsh tool. With no prompt it opens the browser UI. With a prompt it runs headless. Wally registers the model and starts dsh. It does not pass images.

​
Claude Desktop

Claude Desktop is the desktop app. This covers the chat and Cowork tabs. Wally points the app at your model, shows a name the app already knows, and puts the app back when it quits. --restore is for a run that was interrupted before that. New Attach or paste a picture in chat. The same image-capable hosted models take it.

​
OpenClaw

OpenClaw is a personal agent that runs its own runtime. Wally starts tui --local and merges into your openclaw.json, so your wizard state, agents, and gateway token stay. New Image input is on for a model that takes pictures, and off for the rest.

​
Prime Agent

Prime Agent is Prime Intellect’s coding agent. Wally registers our provider and starts it on the model you named. New Image input is on for a model that takes pictures, and off for the rest.

​
DeepSec

DeepSec is Vercel’s open-source agent-powered vulnerability scanner. It maps risky surfaces in a repo, investigates candidates with a coding agent, and writes findings under .deepsec/. Wally does not bundle it. It runs DeepSec 2.3.10 from PATH when that version is installed, otherwise npx --yes deepsec@2.3.10. You need Node 18 or later unless that deepsec is already on PATH. This harness is hosted-only. Sign in with wally account login, or export a Cloud key as WALLY_CLOUD_KEY for a long run. Bare wally deepsec is DeepSec init in the current directory, so run it from a repo you own, not from $HOME. Pass DeepSec’s own command after --. init, setup, process and revalidate get the hosted route. Other subcommands such as status pass through. Choose the model with Wally’s -m; do not pass DeepSec’s own --model on those routed commands.

​
VVAH

VVAH is Visa’s open-source Vulnerability Agentic Harness: an agentic SAST pipeline that surveys a repo, threat-models it, verifies findings, and emits a Markdown and SARIF report. Wally does not bundle it. It runs VVAH 1.5.1. If vvaharness is missing, Wally prints the install command and can pip-install that tag into a venv under your Wally profile. You need Python 3.11 to 3.14 on PATH. This harness is hosted-only. Sign in, or export WALLY_CLOUD_KEY. The command above scans the current directory and does not remediate. Bare wally vvah runs scan --repo . --remediate, which can edit files. Run either from a project directory you are allowed to change, not from $HOME. Pass VVAH’s own command after --. Other commands such as report pass through.

​
Shannon

Shannon is Keygraph’s open-source AI pentester for web apps and APIs. It reads source, then runs real exploits against a live target and reports only issues it can prove. Wally does not bundle it. It runs Shannon 3.4.0 from PATH when that version is installed, otherwise npx --yes @keygraph/shannon@3.4.0. You need Docker running, and Node 18 or later unless that shannon is already on PATH. Linux, macOS, or Windows through WSL2. This harness is hosted-only. Sign in, or export WALLY_CLOUD_KEY for a scan that outlasts a sign-in. With no Shannon command, Wally runs start. Other commands pass through:
A Shannon scan sends attack traffic at the URL you pass. Use a staging or local app you are authorised to test, not production. The longer Shannon guide covers Cloud keys and pointing Shannon at Wally by hand.

​
Anything else

A tool Wally has not been taught about still works. Ask for the endpoint and point the tool at it yourself:
That prints an OpenAI-compatible base URL and holds it open until you stop it. Any client that takes a base URL works against it unchanged.