Capital · Intelligence · Chain
Autonomous capital,engineered on-chain.
TriFi builds and runs the Tokenized Asset Manager — a fleet of autonomous strategy daemons allocating real capital across Solana DeFi. AI makes the calls. Deterministic rails hold the line. We build the same class of system for crypto, blockchain, AI, and web3 teams.
- Fleet
- 14 strategy daemons
- Decision layer
- AI + hard rails
- Settlement
- Solana, continuous
- Posture
- Capital preservation first
Why the name
Three things have to be right before software can be trusted with money on-chain: the capital has to be allocated well, the intelligence making the calls has to be accountable, and the chain underneath has to settle exactly what you intended. TriFi is built at that intersection.
Capital
Allocation, not opinions
A portfolio manager sizes every strategy against live market conditions and each strategy's own realised performance — then writes the budget down as a machine-readable directive the rest of the fleet has to obey.
Intelligence
Models inside a harness
Claude reasons over market structure, pool quality and position health — but it never has the last word. Every model call sits inside retries, fallbacks, cost accounting, and mechanical limits it cannot argue past.
Chain
Settlement you can audit
Every action lands on Solana through routed swaps, protocol integrations and per-strategy wallets, with actual gas and slippage recorded back so the next estimate is calibrated by what really happened.
The Tokenized Asset Manager
A closed loop from market data to settled position.
TAM is TriFi's own system — not a demo. A portfolio manager reads the market, decides the split, and writes directives. Each strategy daemon picks up its directive, runs its own wallet, and executes on-chain. Rails constrain every step, and a watchdog audits the whole fleet from outside it.
01 ingest
Read the market
Price, volatility, funding, flows, yields and pool state pulled from a dozen independent market and protocol feeds.
02 decide
Size the book
The portfolio manager classifies the regime and sets a budget per strategy, bounded by hard per-strategy allocation limits.
03 direct
Write the directive
Decisions become plain JSON on disk — one directive per strategy, re-read at the start of every cycle. No hidden coupling.
04 execute
Run the strategy
Each daemon converges on its budget in its own isolated wallet: enter, rebalance, compound, rotate, or stand down.
05 guard
Hold the line
Stop-losses, drawdown halts, peg-deviation exits, reserve floors and rate-limit backoffs run mechanically, outside the model.
06 observe
Audit and feed back
An out-of-process watchdog checks heartbeats and health; realised costs and PnL calibrate the next cycle's estimates.
Market and protocol inputs
- CoinMarketCap
- Deribit DVOL
- OKX funding + OI
- BTC ETF flows
- Stablecoin flows
- Fear & Greed
- Alt Season Index
- Meteora pools
- Jupiter vaults + lend rates
- DefiLlama TVL
- CoinGecko OHLCV
- Solana DEX volume
The loop runs continuously and unattended. Cycle cadence is per strategy — minutes for mechanical rotation, hours for leveraged carry — and the portfolio manager can trigger an early cycle when the market moves enough to invalidate the current split.
The fleet
Fourteen daemons, one book.
Strategies are deliberately uncorrelated and deliberately isolated. Each runs its own process, its own state file, and its own wallet — so a bad day in one place cannot reach capital anywhere else.
- pm
Portfolio Manager
Reads the market, classifies the regime, and allocates across every other strategy. The sole decision-maker for idle capital, and the only daemon that moves funds between wallets.
- dlmm
Concentrated Liquidity
Provides concentrated liquidity on volatile pairs, earning fees from trading volume in any direction. Trend pre-filters, rebalancing, drawdown halts and a fast-bleed emergency exit manage impermanent loss.
- stable-dlmm
Stable-Pair Liquidity
The same concentrated-liquidity machinery restricted to stablecoin pairs, where the dominant risk is a broken peg rather than a trend. Peg deviation is monitored continuously with a mechanical exit.
- usd-stable-lend
Stablecoin Lending
Parks idle stablecoins at supply yield instead of zero. No impermanent loss, no liquidation risk, withdrawable on demand — the default home for capital when nothing better is on offer.
- usd-stable-arb
Stablecoin Rotation
Pure math, no model. Continuously quotes routes between stablecoins and rotates only when the spread clears execution cost by a configured margin. The fleet's proof that not every problem needs an AI.
- carry3 profiles
Leveraged Carry
Config-driven leveraged carry on lending vaults — staked-SOL, liquidity-provider tokens, and pegged USD pairs. Each profile carries its own leverage ceiling, slippage budget and health-factor floor.
- token-portfolio6 profiles
Token Portfolios
Themed spot books — blue chips, staked SOL, perp LPs, Solana DeFi, DePIN, and major memes — with AI-set allocations wrapped in mechanical stop-losses, take-profits and trailing stops.
- watchdog
Watchdog
Runs outside the fleet and trusts none of it. Checks heartbeats, circuit breakers, carry health factors, drawdowns and state-file corruption, then alerts a human when something needs a human.
The roster is a config file, not a code change — the strategy registry is the single source of truth for allocation, so adding or retiring a strategy is one entry. Strategies that don't suit the current regime simply receive no budget and stand down.
Safety architecture
The model decides. The rails hold.
Every serious question about autonomous finance reduces to one: what happens when the model is wrong? It will be wrong. So the answer has to be structural rather than clever — in TAM, everything that can lose money is deterministic code with tests around it.
Failures stay where they start
Each strategy is its own process, its own state, and its own wallet. A bad route or a stuck position is bounded by that strategy's capital and cannot reach the rest of the book.
Rails are evaluated before the model is asked
Stop-losses, drawdown halts, peg-deviation exits and reserve floors run first, mechanically, every cycle. Whatever the advisor then recommends is clamped against hard limits before anything is signed.
Something outside the fleet is watching it
An independent watchdog reads state and logs from out-of-process and escalates to a human. A daemon that has quietly stopped thinking still gets caught, because nothing asks it to report on itself.
Engineering
We build this class of system for other teams.
The Tokenized Asset Manager is the reference implementation of how we work: production autonomy with real consequences, model reasoning inside a hard harness, and a test suite around anything that touches money. We bring the same engineering to crypto, blockchain, AI, and web3 products.
Autonomous trading & treasury systems
Long-running strategy daemons, capital allocators, execution routing, risk rails and reconciliation harnesses — the machinery that lets software operate a book without a person watching it.
- allocation engines + directive protocols
- mechanical risk rails and circuit breakers
- PnL reconciliation + cost attribution
AI decision layers you can ship
Putting a model in production is mostly everything around the model. We build the harness: structured outputs, retries and fallbacks, cost tracking per call, and deterministic guardrails on every consequential action.
- structured advisors + schema validation
- retry, model fallback, cost accounting
- agent loops with human decision gates
Protocol & DeFi integration
Direct integration with the protocols that actually hold the capital — DEX routing and aggregation, concentrated liquidity, lending and leverage vaults, staking, and the RPC layer underneath it all.
- swap routing + slippage control
- LP, lending and leverage position management
- RPC failover + transaction reliability
Token & web3 product surfaces
The user-facing half: wallet-native front ends, mint and distribution flows, staking and vesting interfaces, and the legal and consent surfaces that have to ship alongside them.
- wallet connection + on-chain UX
- mint, vesting, staking interfaces
- privacy, terms and consent surfaces
On-chain data & observability
You cannot operate what you cannot see. Structured logging, telemetry, cost and slippage attribution, health monitoring and alerting — built in from the start rather than bolted on after the first incident.
- structured logs + rotation + retention
- product analytics + event pipelines
- health checks, alerting, incident routing
Web platforms & content systems
Fast, statically rendered, accessible sites and dashboards on a component-and-content architecture — including the CMS this site runs on, deployed through automated pipelines to global CDNs.
- static rendering + SEO-correct routing
- WCAG 2.2 AA as a build requirement
- CI/CD to CDN, custom domains, analytics
Engagements usually start with one hard problem — an execution path that keeps losing money to slippage, a model that needs guardrails before it can be trusted with anything real, a protocol integration nobody wants to own. That's the right place to start.
Tell us what you're building.
Whether it's an autonomous system that needs to survive contact with real capital, a model that needs a harness before it ships, or a protocol integration nobody wants to own — start with the hard part.