Skip to content
Swytchcode logo Swytchcode logo Swytchcode Docs docs

ZeroClaw

This quickstart sets ZeroClaw up as a personal agent that runs third-party API calls through Swytchcode.

ZeroClaw grants MCP servers and skills to each agent through bundles, so Swytchcode doesn’t change its config for you: swy init prints exactly what to add.

Step 1: Install the Swytchcode CLI

Terminal window
npm install -g swytchcode

Step 2: Create the agent’s project

Terminal window
swy init --operator --editor=zeroclaw --project ~/agent-swy
swy login

This creates the project in sandbox mode, adds the baseline approval policy, and writes the Swytchcode skill to ~/agent-swy/agent-skill/swytchcode-operator/SKILL.md (with always: true, so it stays in the agent’s prompt).

Step 3: Add the skill to a bundle

Terminal window
mkdir -p ~/.zeroclaw/shared/skills/swytchcode
cp -R ~/agent-swy/agent-skill/swytchcode-operator ~/.zeroclaw/shared/skills/swytchcode/

If you set ZEROCLAW_CONFIG_DIR, use that folder instead of ~/.zeroclaw; init prints the right paths.

Step 4: Add the MCP server

Add the block init printed to ~/.zeroclaw/config.toml. It looks like this:

[[mcp.servers]]
name = "swytchcode"
command = "swytchcode"
args = ["mcp", "serve", "--profile", "operator", "--project", "/home/you/agent-swy"]


[mcp_bundles.swytchcode]
servers = ["swytchcode"]


[skill_bundles.swytchcode]

This only defines the server and bundles. No agent gets them until you grant them.

Step 5: Grant it to your agent and restrict it

Give your agent the bundles and a risk profile:

[agents.assistant]
mcp_bundles = ["swytchcode"]     # add to the existing list
skill_bundles = ["swytchcode"]   # add to the existing list
risk_profile = "personal"

Then set up the risk profile. Setting a key replaces ZeroClaw’s whole default list for it, so start from the defaults (zeroclaw config get risk_profiles.<name>.<key>) and add to them:

[risk_profiles.personal]
# ZeroClaw's defaults, plus the project and ~/.swytchcode
forbidden_paths = ["/etc", "/root", "/home", "~/.ssh", "~/.aws", "~/.config", "...", "/home/you/agent-swy", "~/.swytchcode"]
# no shells or interpreters; the default list has python, python3, node, git, npm and pip
allowed_commands = ["ls", "cat", "grep", "head", "tail", "wc", "date"]
# ZeroClaw's defaults, plus the 11 Swytchcode tools (init prints them)
auto_approve = ["file_read", "memory_recall", "...",
  "swytchcode__swytchcode_list", "swytchcode__swytchcode_exec", "..."]
  • forbidden_paths and allowed_commands keep the agent away from policies.json and your credentials. A command that runs code (python3 -c, node -e) can read any path, so leave those out.
  • Without the Swytchcode tools in auto_approve, ZeroClaw asks “[Y]es / [N]o / [A]lways” in the chat before each one. Swytchcode already holds every call that changes something for your approval. swytchcode__* doesn’t work; list each name.
  • Leave workspace_only on (the default).

Then restart the agent’s session.

Step 6: Set the approval channel and check

Set the workspace’s approval channel in app.swytchcode.com (Telegram or Slack, in a private chat with the Swytchcode bot). Then:

Terminal window
swy doctor --agent zeroclaw --project ~/agent-swy

It checks the server, the grant, the skill and the risk profile settings above. It asks ZeroClaw for the values in effect, defaults included.

Step 7: Chat with it

From ZeroClaw’s chat, ask for something that needs a third-party app:

Send an invoice for $200 to Acme through Stripe.

ZeroClaw fetches and enables the Stripe method, sends you a connect link, and the invoice waits for your approval. See What a conversation looks like.