Build Your First Autonomous Workflow
Most trading tools stop working when you close the tab. Cabal does not. This guide walks you through three autonomous workflows -- a price watch, a portfolio review routine, and a tracked wallet -- so your agent keeps running while you are away.
By the end, you will have three things running in the background. Your agent will wake up when conditions change, think about what to do, and either act or surface a recommendation.
Before you start
You need:
- An agent deployed and ready to chat
- At least one model provider connected
- A funded wallet linked to your agent
If you have not done these yet, start with the to get your account set up, then follow to deploy and configure an agent. Come back here once your agent is responding in Chat.
Set a price watch
A price watch tells your agent to monitor a token and wake up when a condition is met. Open Chat with your agent and type:
Watch SOL and alert me if it drops below $170.
That is all it takes. Behind the scenes, Cabal creates a watch that evaluates the condition on a recurring schedule. While the watch is active, you do not need to keep Chat open.
When SOL drops below $170, your agent wakes up on its own. It looks at your positions, checks your standing instructions, and decides what to do -- it might propose buying the dip, suggest trimming exposure, or just log the event for your review. You see a summary in your main chat, and you can open the full thread to read the agent's reasoning.
This is not a dumb price alert that pings your phone. The agent thinks about context before deciding what to surface.
You can set watches on any supported token, and conditions can be above or below a target price.
Schedule a portfolio review
Watches react to specific events. Routines run on a schedule regardless of what the market is doing. In Chat:
Create a routine that reviews my positions every 30 minutes.
Cabal sets up a recurring routine. Every 30 minutes, your agent checks your balances, open positions, and recent activity, then thinks about whether anything needs your attention.
To make the routine useful, give it specific instructions. You can include these in the same message or update them later:
Check if any position has moved more than 10% since last review. Flag anything that conflicts with my standing instructions.
Standing instructions are the persistent preferences you set in your agent's system prompt -- things like "never hold more than 20% in a single token" or "avoid low-liquidity pairs." The routine checks your portfolio against these instructions every time it runs.
Cadence options:
- Every 5 minutes -- for active markets or tight strategies
- Every 15 minutes -- a good balance for most setups
- Every 30 minutes -- the default, suitable for swing positions
- Every 60 minutes -- for longer-horizon strategies
If the agent finds something worth flagging, you get a notification.
Track your first wallet
Wallet tracking turns on-chain activity into actionable signals. In Chat:
Track wallet
7vfCXTUXx5W6Z4VEAnX8VfVCK41sJLdnLhMrDXe4Y35non Solana. Label it "Alpha caller". Only flag transactions above $10K.
Cabal starts monitoring the wallet. The two things that matter most here are the label (so you can identify signals at a glance) and the instructions (so your agent knows what this wallet means to you).
"Only flag transactions above $10K" tells the agent to filter out noise. You can be more specific: "This wallet front-runs new token launches. Treat large swaps into unknown tokens as early signals." Without instructions, the agent still sees the raw data but has no context for what the source means.
You choose how signals reach you:
Instant -- your agent reacts to each transaction as it happens. Best for high-conviction wallets where speed matters.
Batched into a routine -- signals accumulate and arrive together on a schedule. Your agent sees all the activity at once and can spot patterns instead of reacting twelve separate times. To use this, set up a routine first (as described above), then route signals through it:
Deliver signals from "Alpha caller" through my portfolio review routine.
Either way, the agent does not just say "wallet bought X." It reasons about whether the activity matches your strategy -- using both your per-wallet instructions ("only flag above $10K") and your standing instructions ("never chase low-liquidity tokens"). That reasoning is what separates this from a block explorer.
See the results
After your first night with these workflows running, check what happened. Open your chat -- you should see summaries from the overnight runs. Or just ask your agent:
What happened while I was away?
You should see things like:
- Your price watch fired overnight and the agent wrote up its analysis in a branched thread
- Your portfolio review routine ran on schedule and flagged a position that moved 15%
- A tracked wallet made a swap and the agent explained why it did or did not match your strategy
If your agent proposed any trades while in suggest-only mode, those show up as pending approvals you can act on from the chat.
Turn on autopilot
Up to this point, your agent has been in suggest-only mode. It can research, analyze, and propose trades, but every execution needs your manual approval. That is the right starting point -- it lets you see how the agent thinks before you trust it to act.
When you are ready, you can move to autonomous guardrailed mode. In this mode, your agent executes trades on its own, as long as every trade passes your guardrail checks. Guardrails are deterministic checks -- max trade size, max daily volume, max slippage, minimum liquidity -- that run outside the agent. They are simple comparisons: is the trade larger than your limit? Yes or no. The agent cannot see these checks, modify them, or talk its way past them.
To make the switch:
- Open your agent's settings page.
- Find the control profile section.
- Change from Suggest Only to Autonomous Guardrailed.
- Save.
That is it. Your existing watches, routines, and signal sources do not change. They keep running on the same schedules and conditions. The difference is that when your agent decides a trade is worth making, it can execute within your guardrails instead of waiting for you to approve.
Before switching, review your guardrail settings to make sure the limits match your risk tolerance. See for the full reference on every setting and its default value.
What to try next
You have three workflows running. Here are some directions to explore:
- -- get more from your tracked wallets by letting your agent mirror trades with its own analysis layer
- -- tune risk limits to match your specific strategy instead of relying on defaults
- -- approve trades and interact with your agent from Telegram, so you do not need to open the dashboard