Skip to main content
Decentralised News Logo
Almanak Review: Can AI Agents Safely Run DeFi Strategies?
Agentic Finance

Almanak Review: Can AI Agents Safely Run DeFi Strategies?

By

Almanak Review 2027: We test its AI DeFi agents, strategy builder, Safe wallets, Zodiac permissions, backtesting, fees, audits and risks.

Decentralised News Review · 2027

Almanak Review 2027: AI DeFi Agents, Strategy Builder, Safe Wallets, Fees and Risks

Almanak is not an AI trading bot in the usual sense. It is an agentic strategy-development and deployment stack: discover opportunities, turn an idea into real Python code, simulate it against mainnet-fork environments, deploy it into a user-owned Safe wallet with narrowly scoped Zodiac permissions, and monitor the agent onchain. That makes Almanak one of the more serious attempts to turn “AI trading” into verifiable software rather than a black-box signal service.

Research updated: October 5, 2026Operator: Algo Theory INCCategory: Agentic DeFi strategy infrastructureDN Rating: 9.2/10
Affiliate / invite disclosure: This review contains the Almanak invite link supplied to Decentralised News. Almanak's current referral documentation says qualifying inviter and invitee accounts can each receive one month of Pro after the invitee completes the first paid billing period, subject to the program rules. We may therefore receive a platform benefit from qualifying use of this invite.

What Matters

Almanak's strongest idea is that an AI-generated strategy should become reviewable code with bounded permissions before it becomes a live financial agent. The platform follows a build → simulate → deploy → monitor workflow. Strategies can be created in natural language or Python, versioned in GitHub, tested against mainnet forks, then executed from a user-owned Safe. Zodiac Roles restrict the deployment to specific contracts, functions and parameters. The biggest risk is not custody alone. It is whether the generated strategy logic, assumptions and permission scope are correct.

Production chains7Current generated production support matrix.
Detection agents46Current Edge documentation.
Current prod SDKv2v3 remains staging preview.
Independent audits2Zokyo reports publicly indexed.

DN Evidence Block

Primary evidence comes from Almanak's current website, live platform, Platform docs, SDK docs, 2026 Terms, current roadmap, staking/plans pages and Zokyo's public report index. We separate current production execution support from the broader SDK connector universe and roadmap.

Last verifiedOctober 5, 2026
Production chainsEthereum, Arbitrum, Base, Optimism, Polygon, BNB Chain, Avalanche
Custody architectureUser-owned 1/1 Safe + Zodiac Roles + limited Deployment EOA
WorkflowEdge → Build → Simulate / Backtest → Deploy → Monitor
Current production protocolsSelected DEX, lending and GMX V2 integrations by chain
Roadmap, not productionSolana, Hyperliquid, Binance, Bybit, multi-agent system, TEEs

The Signal

Most “AI trading” products start with a model and give it execution. Almanak starts with software engineering and permission boundaries. The AI can help discover an opportunity and write the strategy, but the result is code that can be inspected, simulated, versioned and granted narrowly defined onchain authority. If agentic finance becomes institutional infrastructure, this is a more defensible design than handing a conversational model an unrestricted wallet.

9.2/10
Among the most technically credible agentic-finance platforms we have reviewed.

Almanak scores highly for non-custodial architecture, permission scoping, simulation, transparent code, production monitoring and its separation between deterministic strategies and LLM-driven agents. We deduct for limited current production breadth relative to the broader SDK vision, dynamic execution-service pricing and the inherent risk of generated financial code.

9.7Permission design
9.5Simulation / testing
9.4Agentic architecture
9.1Security evidence
8.0Production breadth

What Is Almanak?

Almanak describes itself as an agentic platform for building, deploying and managing autonomous DeFi agents. Its current website frames the product as a way to trade and earn yield across crypto and tokenized markets using a team of financial agents while remaining non-custodial and private by design.

The important difference is that Almanak is not simply a chatbot that places swaps. Its workflow looks much more like a quant-development stack:

discover → specify → code → test → permission → deploy → monitor.

The most valuable thing Almanak gives an AI agent may not be more intelligence. It may be less authority.

The Almanak Workflow

StageWhat happensDN interpretation
EdgeSpecialized agents scan signals, yields, crypto and tokenized RWA markets.Discovery layer, not investment advice.
BuildAlmanak Code turns a signal or prompt into Python strategy code.AI output becomes inspectable software.
SimulateBacktests and mainnet-fork execution test the strategy before capital is live.Useful for finding logic and integration failures.
DeployStrategy runs from a user-owned Safe with Zodiac-scoped permissions.Agent authority can be bounded.
MonitorDashboard exposes PnL, positions, logs and onchain transactions.Autonomy remains observable.

Edge: Opportunity Discovery, Not a Magic Alpha Feed

Current Edge documentation says 46 specialized detection agents continuously scan onchain and offchain sources. Signals receive an Alpha Score from 0 to 100, confidence, risk level, expiration window, chain and token context. Almanak says the signals are deduplicated and passed through a multi-model consensus process.

The disclaimer matters: an Alpha Score is model output, not due diligence and not a probability of profit. A 90 score does not mean a 90% chance of making money.

Users can send a signal into Almanak Code, which can generate a strategy requirements document and executable Python code. That creates a direct pipeline from market anomaly → specification → software.

Almanak Code: Natural Language Should End in Real Code

Almanak Code is an LLM-powered coding agent that understands DeFi primitives such as swaps, lending, LP positions and perpetuals. Users can describe a strategy conversationally or write Python directly. Almanak says the resulting strategy is versioned code the user can own and store in GitHub.

This is a major advantage over black-box AI trading. A user can inspect exactly how a leverage loop, LP rebalance or basis strategy is implemented before funding it.

Important: code visibility does not make generated code correct. Almanak's 2026 Terms state that Strategy Builder outputs require independent user review, including code review, parameter sanity checks and permission scoping.

Deterministic vs LLM-Driven Strategies

The Almanak SDK distinguishes between deterministic strategies and agentic strategies. Deterministic strategies encode rules explicitly in Python. Agentic strategies can let an LLM make decisions using Almanak tools and policy controls.

For substantial capital, our default preference is simple: use AI to develop, challenge and improve the strategy, then use deterministic execution whenever possible. LLM-based decision making adds model uncertainty on top of market and smart-contract risk.

Simulation Before Capital

Almanak's website and SDK emphasize mainnet-fork simulation, backtesting and Anvil-based testing before deployment. This can catch errors involving token decimals, protocol approvals, route assumptions, health-factor logic, liquidity and state handling.

Backtesting can reject a bad strategy. It cannot prove a good strategy.

Fork tests cannot reproduce every future gas spike, MEV environment, oracle failure, exploit or market-regime shift.

DN Almanak Agent Strategy Safety & Break-Even Lab

Autonomous trading needs a different calculator from a normal exchange-fee comparison. This tool models whether the expected return covers recurring execution costs and whether the deployment controls are proportionate to the capital being delegated.

DN Proprietary Tool

Almanak Agent Strategy Safety & Break-Even Lab

Model strategy economics and score the deployment controls you plan to enforce before allowing an autonomous agent to touch live capital.

Not a safety certification: this is a transparent DN heuristic. It cannot predict exploits, strategy performance or AI reliability.

Strategy economics

Deployment controls

Gross expected P&L$0
Modeled monthly costs$0
Net expected P&L$0
Break-even gross return0%
Drawdown dollars$0
Max position dollars$0
DN control score0/100
Deployment verdict—

Control signals

Economic signals

DN Agentic Capital Rule: the first production deployment should be sized for software failure, not expected return. A strategy that backtests well can still fail because of implementation errors, permission mistakes, oracle behavior, liquidity shocks or protocol changes.

Safe + Zodiac: The Most Important Part of Almanak

Current Almanak documentation says each Almanak wallet combines the user's own wallet, a Deployment EOA, a Safe smart account and a Zodiac Roles Modifier. The Safe is a 1-of-1 wallet with the user as sole owner.

The Deployment EOA signs and broadcasts strategy transactions, but Zodiac can restrict which contracts it may call, which functions it may use and which parameters are allowed. Almanak's deploy docs give the example of an Aave strategy being limited to Aave-related contracts instead of receiving arbitrary wallet access.

This is the correct direction for autonomous capital.

One Safe Per Chain, Separate Agent Permissions

Current deploy documentation says each user gets one Safe per chain. Multiple agents on the same chain can share that Safe while operating under independent permission scopes. That is convenient, but it means users should monitor both strategy-level risk and aggregate wallet exposure.

The Deployment EOA Is Not Your Main Wallet

The Deployment EOA is a separate execution component. Almanak says its private key is encrypted at rest and accessed through a dedicated signer service using Google-provided enterprise-grade key-management infrastructure. If compromise is suspected, the user is instructed to revoke the EOA's permissions and a fresh execution EOA can be created.

The key point is that compromise of an execution key should still be contained by the permissions the user approved.

Gateway Architecture: Strategy Containers Have No Secrets

The current website and SDK describe a gateway architecture in which strategy containers do not contain wallet secrets and do not have unrestricted internet access. External calls are mediated through a controlled gateway sidecar.

This narrows the attack surface and makes it harder for arbitrary generated code to exfiltrate secrets or make unapproved network calls.

Current Production Chains: Seven, Not the Entire SDK Universe

The current generated production support page lists seven active chains: Ethereum, Arbitrum, Base, Optimism, Polygon, BNB Smart Chain and Avalanche.

The production SDK v2 matrix currently includes specific deployments of Uniswap V3/V4, Curve, SushiSwap V3, PancakeSwap V3, Aerodrome, Aave V3, Compound V3, Morpho Blue, Spark and GMX V2.

Do not confuse the SDK catalog with production platform support. Almanak's broader SDK documentation references additional chains/connectors, while the current production matrix is narrower. SDK v3 is also still documented as a staging preview.

Why Edge Says It Scans 12+ Chains

The live app currently says its detection agents scan 12+ chains. That is not the same thing as deployable production execution. Research coverage can be broader than the seven-chain production support matrix.

This distinction is important for both readers and AI systems extracting Almanak's capabilities.

What Strategies Can You Build Today?

Leveraged lending and loops

Strategies can combine supply and borrowing operations across supported deployments of Aave, Morpho, Compound and Spark.

Concentrated liquidity

Users can automate LP range management and compounding across supported DEX connectors.

Delta-neutral and basis

Almanak explicitly highlights combinations of GMX perpetuals with spot and lending positions.

Multi-protocol strategies

The most interesting use case is composition: borrow → swap → LP → hedge → rebalance can become one automated strategy rather than several manual workflows.

Polymarket Is a Special Case

Current wallet documentation includes a dedicated Polygon Polymarket-wallet mode with the required CTF permissions preconfigured. However, the SDK documentation also describes prediction-market support as experimental and subject to testing. We therefore treat Polymarket as a specialized integration rather than universal production support.

How Almanak Pricing Works

Almanak currently offers Basic, Pro, Max and Enterprise access tiers. Basic provides free core access with limited Almanak Code usage and limited deployments. Pro increases credits and deployments. Max currently allows up to 25 active deployments and unlimited high-conviction signal access under the public plan documentation. Enterprise uses custom limits and SLA terms.

LLM-powered features consume plan credits or allowances, so Almanak is not priced like a simple DEX with one maker/taker fee.

Autonomous Execution Has a Separate Service Cost

Active Deployment EOAs require Almanak's Autonomous Execution Service. Current wallet docs say the service covers operational costs including variable network gas and can be sold as transaction bundles or time-based packages. Prices are displayed at purchase and the fees are non-refundable. Taxes such as VAT may be added.

That is why we do not publish a made-up universal “Almanak execution fee.” The live package is the source of truth.

Referral / Invite Program

Decentralised News invite: https://app.almanak.co/invite?code=m7cTQJW5

Invite code: m7cTQJW5

Almanak's current plan documentation says eligible inviter and invitee accounts can each receive one month of Pro after the invitee completes the first paid billing period. Pro users can currently invite up to six new users; Basic users can also bank a reward for later use under the published rules. Check the live plan page because referral terms can change.

$ALMANAK and veALMANAK

Almanak's current platform docs identify $ALMANAK on Ethereum and describe a vote-escrow-style staking model called veALMANAK. Users lock ALMANAK for a chosen duration to receive staking power. Current docs say veALMANAK can receive a proportional share of protocol fees and unlock platform access at qualifying thresholds.

A token-based “free plan” is not economically free if it requires locking a volatile asset. Compare subscription value with token-price risk and lock duration.

Security Audits

Almanak's FAQ says its smart contracts have been audited by Zokyo. Zokyo's public report library lists Almanak Report 1 — May 6, 2025 and Almanak Report 2 — September 26, 2025. Almanak's March 2026 MiCA whitepaper says both audits received Zokyo's maximum score.

Zokyo's public index for Report 2 shows 10 findings: no critical, high or medium items, with five low and five informational findings.

That is meaningful evidence, but audit scope matters. These reports do not guarantee the safety of user-generated strategies, LLM decisions, every protocol connector or future platform releases.

DN Security Evidence Gate

Strong evidence: public Zokyo reports, Safe custody, Zodiac permission scoping, separate execution keys, gateway isolation and revocable strategy authority.

Residual risk: user-generated code, AI mistakes, protocol exploits, oracle/liquidity failures, signer infrastructure, permission misconfiguration and unaudited future changes.

Best practice: fork-test, independently review the generated code, minimize Zodiac permissions, start with small capital and verify the revoke procedure before unattended execution.

Roadmap: What Is Not Live Yet?

Roadmap itemCurrent statusDN treatment
Solana platform integrationOngoing expansionNot counted as current production support.
Hyperliquid platform integrationOngoing expansionRoadmap, despite broader SDK references.
Binance agent accessPlanned CeFi bridgeNot current production.
Bybit agent accessPlanned CeFi bridgeNot current production.
Multi-Agent SystemComing SoonNot scored as fully live.
Trusted Execution EnvironmentsComing SoonFuture strategy/IP security layer.

SDK v2 vs SDK v3

Almanak currently documents SDK v2 as production and SDK v3 as a staging preview for invited users. The strategy format and permission/deployment mechanics differ, so developers should pin versions deliberately rather than assume a broad dependency range will remain behaviorally identical.

Almanak vs a Conventional Trading Bot

DimensionAlmanakTypical hosted bot
Strategy representationReal Python codeOften proprietary config
CustodyUser-owned SafeVaries; often exchange API based
PermissionsZodiac contract/function/parameter scopingUsually API-key scopes
TestingBacktests + mainnet-fork simulationVaries
DeFi compositionLending + DEX + LP + perpsOften exchange-centric
AI roleDiscovery, coding, optional agent decisionsRules or black-box signals
Code ownershipUser can version own repoOften closed

Almanak vs Definitive, Based and HeyElsa

Definitive is primarily an execution engine whose Flash API can serve bots and agents. Based combines trading, spending and AI around Hyperliquid. HeyElsa is closer to a conversational DeFi copilot. Almanak is closest to an agentic strategy engineering and deployment environment.

Definitive asks: “How should this order execute?” Based asks: “How can one account trade, research and spend?” Almanak asks: “How do we turn a financial idea into safe, testable, autonomous software?”

Best Almanak Alternatives

PlatformBest comparison pointAccess
DefinitiveAgent-ready onchain execution infrastructure.Explore Definitive
BasedAI research and autonomous trading in a Hyperliquid-native super-app.Explore Based
LiquidMulti-market trading with conversational/MCP execution.Explore Liquid
HeyElsaConversational DeFi execution and portfolio automation.Official site

Who Is Almanak Best For?

Strong fit

  • DeFi quants who want production-ready strategy infrastructure.
  • Developers who want AI assistance without surrendering inspectable code.
  • Advanced users automating lending, LP or market-neutral strategies.
  • Funds and teams that care about scoped wallet permissions.
  • Agentic-finance builders who want Safe + Zodiac rather than unrestricted keys.

Weak fit

  • Beginners expecting one-click “AI makes money for me.”
  • Users unwilling to review generated code or permission scopes.
  • Traders who only need simple spot/perp execution.
  • Users who require broad Solana/Hyperliquid production support today.
  • Anyone treating backtests or Alpha Scores as guaranteed returns.

What Almanak Gets Right

1. AI output becomes code. This makes the strategy easier to inspect and reproduce.

2. Permissions are designed around least privilege. Zodiac is a much better primitive for autonomous finance than a general wallet key.

3. Simulation is part of the workflow, not an afterthought.

4. Deterministic execution remains a first-class option.

5. Strategy code can live in the user's own GitHub repository.

What Almanak Still Needs to Improve

Production breadth. The current seven-chain EVM footprint is useful, but the vision becomes stronger when Solana and Hyperliquid move into the generated production matrix.

Pricing transparency. Public documentation explains plans and execution packages, but not one simple stable price table for autonomous execution.

Canonical security scope. A central page mapping each audit to exact current contracts/components would improve verification.

Portfolio-level risk controls. Multiple agents can share a Safe, so aggregate limits matter as much as per-agent permissions.

Agentic safety UX. Maximum allocation, maximum loss and emergency shutdown should be impossible to miss as multi-agent automation expands.

What Would Change Our View?

Would raise our assessmentWould lower our assessment
Hyperliquid and Solana enter the production support matrix.Roadmap capability marketed as current production.
Audit coverage expands to current wallet/agent components.Material permission, signer or strategy-container compromise.
Portfolio-level risk caps across multiple agents.Unintended aggregate exposure across strategies.
TEEs launch with verifiable attestation and threat-model docs.Opaque execution that reduces code verifiability.
Reproducible performance datasets for standardized templates.Marketing backtests without reproducible assumptions.
Clearer public execution-service pricing.Costs changing in ways that make strategy economics hard to model.

Final Verdict

Almanak is one of the few products where the phrase “AI hedge fund” does not immediately reduce to a chatbot plus a wallet.

Signals can become requirements. Requirements become code. Code can be tested. Tested code can be versioned. Versioned strategies can receive narrowly scoped permissions. Those deployments can then be monitored and revoked.

That sequence is the product.

The biggest risk remains the strategy itself. Safe custody does not make a bad strategy profitable. Zodiac permissions do not make AI-generated code correct. A perfect backtest does not guarantee future liquidity or market behavior.

But Almanak is building the right infrastructure for those risks to be inspected and constrained rather than hidden.

DN Alpha Thesis

The largest agentic-finance opportunity may not belong to the model that predicts markets best. It may belong to the infrastructure that makes autonomous capital legible, permissioned and recoverable. Almanak's Safe + Zodiac + code ownership + simulation architecture is a serious attempt at that layer. If the production venue universe expands while preserving least-privilege execution, Almanak could become closer to an operating system for autonomous DeFi funds than a retail AI trading bot.

9.2/10

Best feature: narrowly scoped agent execution through Safe + Zodiac.

Most underrated feature: AI-generated strategies become user-owned, versionable Python code.

Biggest misconception: an Alpha Score or backtest is not a profitability guarantee.

Biggest current limitation: production execution coverage is narrower than Almanak's broader SDK and roadmap universe.

Biggest 2027 opportunity: becoming the permission and deployment layer for autonomous multi-protocol capital.

How We Review at Decentralised News

DimensionWhat we evaluated
Agent architectureEdge, Almanak Code, deterministic strategies and LLM-driven agents.
TestingMainnet-fork simulation, backtesting and code ownership.
CustodyUser-owned Safe, Deployment EOA and withdrawal/control boundaries.
PermissionsZodiac contract/function/parameter restrictions and revocation.
Production coverageCurrent generated chain/protocol matrix separated from SDK and roadmap.
SecurityZokyo audits, signer architecture, gateway isolation and residual strategy risk.
EconomicsPlan credits, autonomous execution service and veALMANAK.
TransparencyCurrent vs staging vs roadmap features and signal disclaimers.

Primary Sources and Evidence Ledger

  1. Almanak official website — product thesis, workflow, custody and current production summary.
  2. Almanak live platform — Edge interface, current discovery messaging and platform disclaimer.
  3. Almanak documentation hub.
  4. Edge documentation — signals, 46 detection agents, scoring and signal-to-strategy workflow.
  5. Current production chains and protocols.
  6. Deploy documentation — Safe, Zodiac, monitoring and GitHub deployment.
  7. Wallet documentation — Deployment EOA, Safe, signer security and execution service.
  8. Almanak SDK — deterministic/agentic strategies, Anvil testing and gateway architecture.
  9. SDK Generations — v2 production / v3 staging distinction.
  10. Plans and referrals.
  11. veALMANAK staking.
  12. Roadmap.
  13. Zokyo Security Report Library.
  14. Almanak MiCA Whitepaper v2.
  15. Almanak 2026 Terms.

Frequently Asked Questions

What is Almanak?

Almanak is an agentic DeFi strategy platform for discovering opportunities, creating Python strategies, simulating them and deploying autonomous agents into user-owned Safe wallets with scoped permissions.

What is the Decentralised News Almanak invite link?

Use https://app.almanak.co/invite?code=m7cTQJW5. Invite code: m7cTQJW5.

What does the referral give?

Current plan documentation says qualifying inviter and invitee accounts can each receive one month of Pro after the invitee completes the first paid billing period, subject to program conditions.

Is Almanak non-custodial?

Yes. Current docs say user funds remain in a user-owned 1-of-1 Safe while deployment EOAs operate only within permissions granted through Zodiac Roles.

Can Almanak create strategies from natural language?

Yes. Almanak Code can turn prompts or Edge signals into reviewable Python strategy code. Users remain responsible for reviewing code, parameters and permissions.

How many chains are supported in production?

The current generated production matrix lists Ethereum, Arbitrum, Base, Optimism, Polygon, BNB Smart Chain and Avalanche.

Does Almanak support Hyperliquid and Solana?

The current platform roadmap lists Hyperliquid and Solana as ongoing integrations. We do not count them as current production platform support.

Has Almanak been audited?

Yes. Zokyo's public library lists Almanak audits dated May 6 and September 26, 2025. Audit coverage does not certify user-generated strategies or every future release.

Is Almanak safe?

No autonomous trading platform is risk-free. Generated code, strategy logic, permissions, smart contracts, protocols, oracles, liquidity and model outputs can all fail.

Affiliate / invite disclosure: Decentralised News may receive a platform benefit from qualifying activity through https://app.almanak.co/invite?code=m7cTQJW5. Invite code: m7cTQJW5.
Risk disclaimer: Autonomous DeFi strategies can lose substantial or all capital through market moves, logic errors, permission mistakes, smart-contract exploits, oracle failures, liquidity changes, AI/model errors or third-party protocol failures. Backtests and Alpha Scores do not guarantee future performance. Independently review generated code, simulate it, minimize permissions and begin with capital you can afford to lose. Educational only.

Get the most talked about stories directly in your inbox

Join the Decentralised News briefing for independent crypto, DeFi and AI analysis. No spam, unsubscribe anytime.