Agent decision
Bot receives a valid quote and fires the buy path.
For Solana Trading Bots
Your bot made a bad trade.
Find out exactly why.
Mortem traces every agent decision against real market conditions, shows you the exact moment the strategy broke, and generates the code fix to prevent repeats.
Market event
Spread widens, liquidity shifts, and the route age exceeds the safe window.
Bad outcome
Execution lands on a worse entry than the strategy assumed.
Built for Solana
Traces agent payloads + onchain actions
Jupiter & Pyth market data built-in
The problem
Your bot is live. It's losing money. And you have no idea why.
Standard logs don't tell you if the entry was bad because volatility spiked, liquidity dried up, or the strategy signal was stale. You're flying blind between the decision and the loss.
The Evidence Chain
Follow the failure from payload to patch.
Instrument Once
Drop the Mortem SDK into your agent. One setup, zero overhead. Every decision your bot makes starts logging.
The Moment It Went Wrong
Mortem aligns your agent's payload, the onchain transaction, and live market data on a single timeline. You see exactly when the decision diverged from reality.
The Diagnosis
Not a vague summary. A specific claim tied to timestamps, quotes, spreads, and the actual strategy path that fired.
The Fix Prompt
Mortem generates a targeted code-level fix prompt for the exact failure path. Review it, test it, ship it.
buy(signal=alpha_04)
spread widens · quote stale
entry lands worse than model assumed
Three timestamps. One failure chain. No dashboard sprawl.
Chronological Trace View
You're staring at logs and guessing sequence.
Mortem shows a clean, ordered timeline: agent payload first, then the market shift, then the bad execution. Not 200 events. The 3 that mattered.
Pyth move
+2.8%
Quote age
61.2s
Liquidity
fell 34%
Spread
widened 3.4x
Price, spread, liquidity, and quote age pinned to the same moment.
Market Context Anchoring
You know the trade was bad, but you don't know what market condition caused it.
Every diagnosis is grounded in real-time Pyth price data and Jupiter liquidity quotes, not LLM assumptions. If the claim isn't supported by verifiable data, Mortem doesn't make it.
// reject stale routes before execution
if (routeAgeMs > MAX_ROUTE_AGE_MS) return refreshQuote()
// re-check spread before buy path
if (spreadBps > maxSpreadBps) return skipTrade()
One failing branch. One remediation prompt. One next change.
Fix Prompts, Not Essays
You get a nice explanation with no clear action.
Mortem generates a specific, code-scoped fix prompt tied to the failing strategy path, not generic advice. You paste it into your codebase, review the diff, test it, ship it.
intent classified
jupiter.quote + pyth.prices
swap submitted
x402 settled
Agent-native traces instead of generic app telemetry.
Built for Agent Builders
Most observability tools were built for web apps, not autonomous trading logic.
Native SDK for TypeScript agents on Solana. Trace LLM calls, tool invocations, wallet actions, and x402 payments in one unified agent timeline.
FAQ
The questions technical buyers ask first.
Does my bot need to be live to use Mortem?
No. You can replay past failed trades and run analysis on historical traces. Real-time alerting ships in the next version.
What chains and protocols does Mortem support?
Currently Solana-native. Jupiter quotes and Pyth price feeds are built into the market context layer.
Is this just AI summarizing my logs?
No. Deterministic checks run first for payload structure, market deviation, and execution timing. The LLM explains conclusions already anchored in verifiable facts.
Who is this built for?
TypeScript agent builders and Solana bot operators who run live trading strategies and need to debug and improve execution quality fast.
What does a fix prompt actually look like?
It's a code-scoped instruction targeting the strategy path that failed, not a paragraph of suggestions. You review the proposed diff, run a backtest if needed, and ship only when you're confident.
Is my trade data private?
Yes. Traces are scoped to your wallet and agent. Nothing is shared or used for model training.
Final call
Stop debugging in the dark.
Bring your worst trade. We'll show you exactly what happened.
No credit card. No lengthy setup. Just the SDK and your next bad trade.