5. Context & Triggers
Agents in Hypaw Terminal are powered by live context — structured real-time data streams delivered through the Model Context Protocol (MCP). Every action an agent takes is based on trigger logic that continuously evaluates this context.
This section explains how context works, what types of data are available, and how to define effective triggers.
What is Context?
Context refers to the real-time signals that flow into an agent — things like price movement, wallet activity, trending topics, or sentiment surges.
MCP collects and delivers this data in a unified format. Agents subscribe only to the types of context they’re configured to care about. This keeps them fast, focused, and interpretable.
You don’t write queries or scrape APIs — you simply pick which inputs your agent should watch.
Types of Context Available
Market
Real-time trading data from Hyperliquid and other exchanges
Price, volume, volatility, open interest
Social
Activity across X (Twitter), Telegram, Discord
Mention count, keyword trends, user clusters
Wallet
Onchain token movements and wallet behaviors
Transfers, deposits, label-based tracking
Governance
DAO proposal status and outcomes
Votes, executions, proposal metadata
Meta
Agent performance, cooldowns, trigger frequency
Last trigger time, trigger count, active status
Each agent may subscribe to one or more of these feeds, depending on its purpose.
How Triggers Work
Every agent in Terminal is driven by a trigger condition — a logic block that decides when the agent should act. The condition continuously evaluates incoming context, and if the rule is met, the agent executes its defined action.
Example:
“If $TOKEN is mentioned 500+ times on X AND price increases 4% in 15 minutes → post a tweet.”
This trigger combines two separate signals:
X_mentions >= 500
price_delta_15m >= 4%
The logic engine handles this check automatically, as long as the context is correctly defined in the agent config.
Trigger Configuration Options
You can customize trigger behavior using the following parameters:
Thresholds
Set min/max values for any context input (e.g. mentions > 300)
Time Windows
Define rolling intervals for measurement (e.g. last 30 mins)
Combinators
Use AND / OR logic between multiple conditions
Cooldowns
Limit how frequently the trigger can fire
Caps
Set maximum number of actions per hour/day
Force Trigger
Optionally test your logic manually from the dashboard
Triggers are fully stateless by default — they only evaluate based on the current context and don’t store memory unless specified via cooldowns or counters.
Real-Time Evaluation
Once an agent is live:
MCP pushes new context to the runtime every few seconds
The trigger logic is evaluated immediately
If true, the action is fired
If false, the agent continues listening
All evaluations and actions are logged in the dashboard. You can inspect what data was used and why an action was (or wasn’t) triggered.
Next: 6. Execution Logic → Learn what happens when a trigger fires, how actions are handled, and how strategies connect to Hyperliquid or other endpoints.
Last updated