Skip to content
Keele Labs
Product direction / August 2026

Roadmap

A gated path from EVM research to user-controlled execution.

This roadmap describes Keele Labs' current direction, not a promise of delivery or timing. Scope and sequence may change as we learn.

IN PROGRESS COMPLETED PLANNED
00
CompletedPHASE 00

Decisions & foundations

Non-custodial, EVM-first system design

Documented the custody model, initial EVM chain scope, allowlisted token universe, strategy horizon, reliability definition, and legal posture before building execution features.

ESTIMATE
2–3 weeks
EXIT GATE
Six core decisions documented and signed off.
In progressPHASE 01

Market analysis engine

Building a validated EVM market-state engine — no signals yet

Ingest and retain allowlisted EVM pool state, prices, flows, liquidity, token-risk data, and feature-engineered market context. AI reasoning is grounded in compact evidence, not raw candles or price predictions.

ESTIMATE
10–12 weeks
EXIT GATE
Versioned analysis schema, 99% data-collection uptime over four weeks, and clean raw snapshots.
02
PlannedPHASE 02

Simulation & evaluation

Replay, cost modeling, and paper trading

Replay raw market snapshots through the production feature pipeline, model realistic gas, fees, slippage, and MEV costs, then run a continuously tracked virtual portfolio before real-money execution.

ESTIMATE
6–8 weeks
EXIT GATE
Replay determinism proven and cost model within 15% of reference swap costs.
03
PlannedPHASE 03

Signal engine

Evidence-backed buy and sell candidates

Combine regime-aware hard filters, quantitative scoring, and bounded AI adjudication to produce typed, expiring signal candidates. Staking remains a separate yield workflow in Phase 6.

ESTIMATE
8–10 weeks
EXIT GATE
Out-of-sample paper-trading evaluation must beat the documented baselines net of modeled costs.
04
PlannedPHASE 04

Risk & policy layer

Deterministic controls outside the model

Enforce token and venue allowlists, sizing and loss limits, liquidity and slippage thresholds, rate limits, pre-flight simulation, and circuit breakers. AI never authorizes a transaction.

ESTIMATE
4–5 weeks
EXIT GATE
The adversarial test suite passes and kill-switch behavior is proven end to end.
05
PlannedPHASE 05

User-directed execution

Review-and-sign first; bounded automation later

Integrate established routing, Permit2 approvals, transaction simulation, and MEV-aware delivery. Version one remains user-signed; any later session-key automation is scoped, revocable, and contract-enforced.

ESTIMATE
8–10 weeks
EXIT GATE
Paper trading precedes capped live rollout; realized costs must remain within 15% of the model.
06
PlannedPHASE 06

Yield & staking

Risk-adjusted yield research and execution

Keep yield separate from directional trading. Evaluate sustainable yield sources, lockups, smart-contract exposure, impermanent loss, depeg risk, and reward-token dilution rather than ranking by headline APY.

ESTIMATE
6–8 weeks
EXIT GATE
Risk-adjusted ranking must outperform naive highest-APY selection in evaluation.
07
PlannedPHASE 07

Multi-chain expansion

EVM fan-out first; Solana as a dedicated adapter

Extend the complete analysis, signal, policy, and execution workflow one chain at a time. Additional EVM networks are configuration plus data QA; Solana requires a separate account, routing, and transaction adapter.

ESTIMATE
2–3 weeks per EVM chain / 10–12 weeks for Solana
EXIT GATE
Per-chain data, policy, and execution-parity tests pass before enabling a new network.

How we sequence

Direction before deadline.

Paper trade first

Backtests can overfit. Real-money execution remains gated by forward paper-trading evidence.

Keep AI bounded

Model output is evidence-backed context; deterministic policy decides what may proceed.

Stay non-custodial

Begin with user-signed transactions and treat every new chain and permission surface as its own review.

Want progress notes, not launch hype?

Join the waitlist for occasional product updates.

Join waitlist