# Warden Protocol

> Purpose-built L1 for AI agents with verifiable inference (SPEX). Heavy Qredo/Binance alumni. The 60M agent runs figure includes off-chain counts nobody can verify. Inflated user numbers, token down ~98% from its Feb 2026 peak and still depressed.

- Scores: Freedom 4.5/10 (grade D), Returns 4.8/10 (grade D)
- Category: agent · Token: WARD
- Published: 2026-03-06 · Updated: 2026-10-15
- Canonical: https://ownyourmind.ai/projects/warden/
- Data (scores + integrity verdicts, JSON): https://ownyourmind.ai/api/projects/warden.json
- Cite as: Own Your Mind (ownyourmind.ai), licence CC BY 4.0. Link the canonical URL.

Charts and interactive widgets are omitted here; their data is in the JSON above.

---

## What it does

Warden Protocol is a purpose-built Layer-1 blockchain for AI agents in Web3. It positions itself as the "distribution layer for the agent economy": infrastructure for autonomous AI agents to execute transactions, manage assets, and interact across chains.

Core components:

- **Agentic Wallet** – natural language interface for crypto operations. "The wallet anyone can use – just ask." Agents handle the complexity behind conversational commands.
- **Agent Hub** – an app-store-style marketplace for Web3 AI agents. Pre-built agents include swap agents, trading bots (Kaibot), the Messari research agent, and portfolio management agents.
- **SPEX (Statistical Proof of Execution)** – verifiable AI inference using probabilistic sampling and consensus across validators. Published as an academic paper (arXiv: 2503.18899). Validators independently verify random samples of inference outputs rather than re-running entire computations.
- **AVRs (Asynchronous Verifiable Resources)** – plugins that encapsulate AI functions callable cross-chain. Think microservices for agent capabilities.
- **AI Trading Terminal** – Messari Signals integration for perpetuals and tokenised stocks.
- **Halo** – a peer-to-peer AI inference marketplace on Base, in public alpha since 30 June 2026. Consumers pay operators per request in USDC through an on-chain escrow, and the protocol withholds a fee from the operator payout, 10% at launch per the run-halo README. Warden names Venice.AI, NEAR and 0G as founding inference contributors.

Multi-chain support via IBC (Cosmos origins), Hyperlane, and deBridge. Covers Solana, Ethereum, BNB Chain, Base, Arbitrum, and WardenChain.

The project spawned from work at **Qredo** (which separately became Open Custody Protocol). The team is heavy on Qredo and Binance alumni:

- **David Pinger** (Co-Founder, Ecosystem) – LMU Munich + Pantheon Sorbonne, ex-Uber, ex-Binance, led R&D at Qredo Labs.
- **Antonio Pitasi** (Co-Founder, Protocol) – ex-Tendermint/Ignite, Cosmos SDK expertise.
- **Alex Melikhov** (Co-Founder, Product) – CEO of Equilibrium (EOSDT stablecoin).
- **Jon Heywood** (Co-Founder, Design) – 15+ years tech, ex-Qredo, built SpaceWard.
- **Josh Goodbody** (CEO) – ex-Binance, ex-Qredo.
- **Dr. Michele Dallachiesa** (Chief AI Officer) – ex-Google, ex-NASA.

Entity based in Madrid, Spain. Total funding approximately $5M across three rounds:

- **Angel/seed:** ~$450K (early 2024)
- **Fjord Foundry IDO:** $534K at $460M FDV (April-May 2024)
- **Strategic round:** $4M at $200M valuation (January 2026 from 0G, Messari, Venice.AI)

Token launched 4 February 2026. As of September 2026 it has traded for roughly seven months.

## Value proposition

SPEX is the technical standout. Rather than requiring expensive ZK proofs or running every computation on-chain, SPEX uses statistical sampling: validators independently verify random samples of AI inference outputs, and consensus determines correctness. This is a practical middle ground between full verification and trust-based oracles.

Warden's [2025 report](https://wardenprotocol.org/blog/warden-report-2025) claims 60.4 million AI agent runs, 11.7 million inference proofs, 10.9 million cumulative on-chain transactions, $749K protocol revenue, and $704K agent revenue. We sampled the chain ourselves via the Cosmos RPC endpoint and the chain is live: blocks land every 2.8 seconds and carry about one transaction each, 1.03 across 60 sampled blocks when we re-ran it.

We previously put our own extrapolation, 35-37 million transactions, against Warden's 10.9 million and called the gap unexplained. That comparison was ours to get wrong and we got it wrong: Warden's figure covers the first three months to 31 December 2025, while our extrapolation ran across the whole chain as it stood months later. Different periods are not a discrepancy. The cumulative count is not something we can settle either way, because the public endpoint is pruned and no Cosmos explorer indexes Warden, so treat the 10.9 million as a self-reported figure with no independent check rather than as one we have contradicted.

The 60.4 million "agent runs" figure is harder to verify. It's 5-6x the on-chain transaction count, which means most of those runs are off-chain activity counted by Warden's own infrastructure. No independent analytics platform (DeFiLlama, Artemis, Dune) tracks WardenChain, so we can't confirm the off-chain portion.

That matters, because there's a pattern of inflated metrics here.

**Inflated user numbers.** Warden claimed 15-20 million users. The founders then admitted to purging approximately 6-7 million bots. The remaining "6.5 million real users" is also unverified by any independent source. When a project inflates user counts by 2-3x and then walks them back, every other self-reported metric deserves the same scrutiny.

**~98% decline, and it has not recovered.** WARD launched at TGE on 4 February 2026, peaked the same week, and remains down roughly 98% from that ATH as of September 2026. Fjord Foundry IDO participants are down over 99%. What looked like a launch-month crash is now a sustained seven-month decline.

**Proprietary core module.** The x/async module is proprietary with restricted use. The rest of the codebase is Apache 2.0, but having a core component closed-source undermines the whole open-source pitch.

42% of supply is controlled by insiders (20% core contributors + 22% treasury), with 120M treasury tokens already unlocked at genesis. Add unlimited PoS inflation running at its 10% ceiling, no burn on chain and just $750K of self-reported 2025 revenue, and the deflationary case doesn't hold up.

The "Do-It-For-Me Economy" framing (agents as autonomous economic actors) is compelling as a vision but describes a future that doesn't exist yet.

## Tokenomics

WARD has a genesis supply of 1 billion tokens with unlimited maximum supply due to PoS inflation. The chain's mint module allows 2% to 10% a year, against the 1% to 10% in Warden's [supply docs](https://docs.wardenprotocol.org/ward/supply), and it has run at the 10% ceiling because bonded stake sits near 14% of supply against a 67% target. Approximately 250 million are circulating (25%).

Distribution: core contributors 20% (6-month cliff + 24-month linear), treasury/R&D 22% (120M at genesis + 12-month linear), agent incentives 19% (20M at genesis + 12-month linear), ecosystem/community 12% (20M at genesis + 30-month linear), validators 10% (permanently locked), public goods 5.3%, liquidity 5%, initial airdrop 4.7%, strategic investors 2% (12-month cliff + 12-month linear).

Token utility: gas on WardenChain, PoS staking, governance voting, agent execution fees, operator bonding. Warden's docs describe a 3% community tax on protocol fees that is burned, but the chain's distribution module takes 2% of block rewards into the community pool. Between 1 and 14 October 2026 the pool grew by 53,968 WARD, 2% of the WARD minted, and total supply rose by 2,698,324 WARD, the full amount minted. No WARD was burned in that window.

Turnover runs near half the market cap daily, suggesting active speculation rather than organic demand. Listed on KuCoin, BingX, Bitget, and Binance Alpha; Bybit lists no WARD pair. Available on WardenChain, BNB Chain, and Base. FDV/MCap ratio reflects 25% circulating supply at TGE with significant unlock pressure ahead.

### Halo and the HALO buyback

The Halo buyback that exists buys HALO, a separate token on Base (`0xbbd27C575fB0e113219D610cc787B02Eeff71d42`). Warden's May 2026 roadmap promised an automated WARD buyback funded by a 10% slice of Halo fees, due in August. Warden's 2 September post routes Halo revenue to HALO buybacks, burns and staking instead, and the team's run-halo ledger records cycles as manual until a permissionless `execute()` contract ships.

We checked the one cycle so far against Base. On 2 September 2026 the executing address `0xf3eEE5bBBF234FD6392465B48Fa1c5f44AeEF6F5` bought 355,971.78 HALO with 32,037.00 VIRTUAL of launch-pool trading fees and 3,752.02 HALO with 238.05 USDC of Halo vault fees, added 373,423.86 HALO of launch-pool fees collected in HALO, then burned 219,944.30 HALO through `burn()` and sent 513,203.36 HALO to the staking contract, matching the ledger. HALO's total supply read 999,780,055.70 at Base block 51,726,662 on 24 September, exactly 1 billion less that burn.  Inference fees account for about 0.5% of the 733,147.66 HALO split between burn and staking that cycle, and trading fees on HALO's own launch pool supplied the rest.

We found no path by which Halo revenue reaches WARD today. Halo therefore leaves Token Utility (12/20) and Revenue Sustainability (8/25) unchanged, and the gap between the WARD buyback Warden promised and the HALO buyback it runs goes to the monthly review.

Warden now points to a second revenue source for WARD. Its [Token Terminal](https://wardenprotocol.org/blog/warden-token-terminal-is-live), launched on 25 September 2026, trades tokens, stocks and RWAs across Base, BNB Chain, Robinhood Chain and Arc, and Warden says "Token Terminal revenue goes into buying back $WARD". That is Warden's statement of how the revenue will be used, and this page will count it as a WARD buyback once a transaction shows one.

Warden also says WARD is bridged to Robinhood Chain at `0xeBaD9A8a051C2A283288a71Ea88eF1B7C69BBe3d` ([2 September post](https://wardenprotocol.org/blog/warden-2-0-ward-takes-the-floor)). Warden 2.0, described as one app for every Warden service, hasn't shipped: the 25 September post lists it as "Coming soon".

### Proposal 6: the move to Base

On 10 October 2026 Warden put [its biggest change](https://wardenprotocol.org/blog/ward-enters-its-revenue-era-governance-proposal) to a governance vote. Proposal 6 moves WARD from WardenChain to an ERC-20 on Base, with versions on BNB Chain and Robinhood Chain. Warden says supply will be capped at 2 billion WARD, inflation will fall from 10% to 5% a year and go to stakers only, and product fees will buy WARD and burn it.

**It passed.** The chain's governance module closed the vote at 08:20 UTC on 15 October with 75.9M WARD Yes, 38.7M No, 9.5M Abstain and no veto. Yes took 66.2% of Yes and No combined, and about four-fifths of bonded stake voted.

What the pass does and doesn't do, from our chain reads on 14 and 15 October:

- **The pass executes nothing on chain.** The proposal carries no messages, so it's a mandate. Warden now has to halt the chain at a snapshot, deploy the Base contract and credit balances itself. Its [FAQ](https://wardenprotocol.org/blog/faq-ward-revenue-era-governance-proposal) says it will share the dates once the vote passes, and none were published by 15 October. WardenChain was still producing blocks that afternoon.
- **The cap doesn't exist yet.** The 2B limit will live in a contract that hasn't been deployed, so we keep WARD listed as uncapped until we can read the cap on Base. Supply read 1,065.0M WARD on 14 October, so the cap leaves room for supply to nearly double. At the proposed 5% a year and before any burn, reaching it would take about 13 years.
- **One address carried most of the No.** On 14 October `warden1f47fj2qejqq4u7tj20wyjdk3qyakpu050qy5ky` had voted No with 35.0M WARD staked, nearly a quarter of the 153.2M then bonded. The chain drops individual votes once a proposal closes, so the final split by voter can't be read back.

The proposal post also says WARD has "No VC allocation. No cliff on the calendar." Warden's own [distribution page](https://docs.wardenprotocol.org/ward/distribution) still allocates 2% to strategic round participants on a 12-month cliff from the February 2026 launch, which falls in February 2027.

## How to participate

**Stake WARD.** Secure WardenChain via PoS. Dynamic inflation targets 67% bonded. Technical skill: basic.

**Use the Agentic Wallet.** Natural language crypto operations. Access pre-built agents from Agent Hub. Technical skill: basic.

**Build Agents.** Use Warden Studio to create and deploy AI agents. SPEX for verifiable inference. Technical skill: advanced.

## Honest assessment

## Freedom Score: 45/100

Warden Protocol is an early-stage project (token live since February 2026) with heavy team control and mixed transparency signals.

**Infrastructure Decentralisation: 8/20.** PoS chain with validator set of 34 active validators, and just 2 of them hold enough stake to halt it. WardenChain is brand new. The x/async module is proprietary, so core infrastructure has centralised control. SPEX is novel but unproven at scale. No DeFiLlama presence. (figures as of 2026-10-19)

**Governance Decentralisation: 8/20.** Token-based governance is young, live since February 2026, and it has been used. Six proposals have reached the chain's governance module, and five passed: three upgrades, a transaction-fee increase and, on 15 October 2026, the migration to Base. One did not: "Set Fixed 3% Inflation Rate" was rejected by token holders in February 2026, with opposing stake outweighing support by more than two orders of magnitude.

A defeated monetary-policy proposal is better evidence of live on-chain governance than a run of unopposed upgrades. It is still a thin record. Treasury holds 22% with 120M unlocked at genesis, core contributors hold 20% with vesting, and centralisation around the founding team remains heavy.

**Token Distribution Fairness: 6/15.** Only 4.7% to initial airdrop. 42% controlled by team (20% core contributors + 22% treasury). Strategic investors only 2%, relatively small. Fjord Foundry IDO participants down over 99%. 25% circulating at TGE.

**Censorship Resistance: 7/15.** Claims zk-SNARK integration for privacy but unverified. Permissionless agent deployment is positive. Multi-chain presence reduces single-point failure. Protocol is young and team-controlled; it could realistically restrict agents or modify rules.

**Data Sovereignty: 8/15.** User key management through cryptographic IDs. SPEX provides verifiable AI inference, so users can verify outputs. Cross-chain operation avoids single-network lock-in. Privacy features claimed but implementation status unclear.

**Open Source Transparency: 8/15.** Main repo Apache 2.0 with over 2,000 stars, decent. However, x/async module is proprietary, a meaningful gap. 38 public repos with active development.  Agent-kit is MIT licensed. Team is publicly identifiable and has published an academic paper on SPEX.

## Returns Score: 48/100

**Token Utility: 12/20.** Gas token on WardenChain with direct utility. Agent execution fee payment is a growing demand driver. PoS staking, governance, operator bonding. Gated functionality and subscription tiers. Actual fee revenue is modest (~$750K cumulative for 2025, self-reported).

**Value Accrual: 10/20.** No WARD burn runs on chain: the 2% community tax pays into the community pool. Validator and operator bonding locks supply. Agent fees paid in WARD create buy pressure. Revenue is small (a self-reported $750K for 2025), and unlimited supply with inflation at its 10% ceiling dilutes holders. Trading about 98% below ATH; market isn't convinced.

**Supply Dynamics: 10/20.** 25% circulating at TGE with structured vesting. Core contributor cliff (6 months) with 24-month linear vest. Validator tokens permanently locked, which is positive. Unlimited max supply with PoS inflation (2-10% on chain). 120M treasury tokens unlocked at genesis is concerning. Significant unlock pressure in months 6-12.

**Revenue Sustainability: 8/25.** Warden reports protocol revenue of ~$750K in 2025 and agent revenue of ~$704K, which would show actual economic activity if it held up to an independent check. $1M agent incentive programme to expand adoption. Multiple revenue streams. Revenue is tiny relative to claimed user base. Sustainability unproven and heavily dependent on agent adoption which remains speculative.

**Liquidity & Access: 8/15.** 13 pairs across 11 venues, led by KuCoin, Bitget and MEXC, with BingX and Binance Alpha alongside them; Bybit carries no WARD pair, checked at Bybit. Multi-chain availability. Turnover runs near half the market cap daily, which is active speculation on a thin book. Micro-cap territory means extreme volatility. Down roughly 98% from its TGE ATH, a decline sustained over seven months.

### Quadrant: D (Low Freedom, Low Returns)

Warden Protocol currently sits in Quadrant D. Centralised team control combined with unproven economics and a collapsing token price. Could move to C or B if agent adoption materialises and governance matures.

### Key risks

- **Sustained ~98% decline.** Down roughly 98% from its TGE ATH. IDO investors down over 99%. Live since February 2026 and persistently weak.
- **Inflated user numbers.** Claimed 15-20M users, admitted purging ~6M bots. Actual engagement unclear.
- **42% insider allocation.** Core contributors (20%) + treasury (22%) with substantial tokens already unlocked.
- **Proprietary core module.** x/async module is closed source, undermining open-source positioning.
- **Unlimited supply.** PoS inflation at its 10% ceiling and no burn running on chain. Proposal 6 passed on 15 October 2026 and would cap supply at 2B and halve inflation, once Warden delivers the migration.
- **Young, thin track record.** Token launched February 2026, roughly seven months of trading. Six governance proposals so far, and no battle-testing.
- **Qredo heritage.** Spawned from Qredo, which had its own controversies and rebranded separately. Team is competent but carries baggage.
- **$750K self-reported 2025 revenue.** Tiny relative to claimed users. The market already prices in limited traction.
- **SPEX unaudited.** Novel verifiable inference system with no third-party security audit published.

## Score change log

| Date | Score | Change | Reason |
| - | - | - | - |
| 2026-10-15 | Data | N/A | Proposal 6, the Base migration, passed on 15 October (66.2% Yes of Yes and No). It carries no on-chain messages, so supply type stays uncapped until the Base contract is live. Scores unchanged. |
| 2026-10-14 | Data | N/A | Correction: no WARD burn runs on chain; the 2% community tax accrues to the community pool (supply and pool deltas, 1-14 Oct). Inflation band corrected to 2-10%. Added proposal 6, the Base migration vote. Scores unchanged; Value Accrual flagged for the monthly. |
| 2026-10-04 | Data | N/A | Added Warden Token Terminal (live 25 Sep), whose revenue Warden says funds WARD buybacks (self-reported, not yet seen as a WARD buyback), WARD's Robinhood Chain bridge, and Warden 2.0 as still "Coming soon". Scores unchanged. |
| 2026-09-24 | Score | Freedom 47 to 45 | Infrastructure 10 to 8: 2 of 33 validators can halt the chain, on top of the proprietary x/async module. |
| 2026-09-24 | Data | N/A | Drawdown refreshed from ~94% to ~98% below ATH and IDO loss to over 99% (CoinGecko, 24 Sep); validator count and halt coefficient bound; repo counts and governance record updated; revenue marked self-reported. Scores unchanged. |
| 2026-09-24 | Data | N/A | Added Halo. Its one buyback cycle (2 Sep 2026, checked on Base) bought and burned HALO, a separate token, where the May roadmap had promised a WARD buyback. Scores unchanged; flagged for the monthly. |
| 2026-09-23 | Data | N/A | Removed the withdrawn ~35M transaction estimate from the description and verdict. Scores unchanged. |
| 2026-09-23 | Data | N/A | Integrity records re-cut: 3 claims now quote Warden's own pages verbatim and 4 are our own measurements. Scores unchanged. |
| 2026-09-18 | Data | N/A | Withdrew our own ~35-37M on-chain transaction estimate: it compared a whole-chain extrapolation against Warden's first-quarter figure. The cumulative count is unverifiable from any reachable source; scores unchanged. |
| 2026-09-07 | Data | N/A | WS3b consensus rail: two validators hold more than a third of stake, the lowest halt coefficient in our corpus, with the largest near 29% and only about 14% of supply bonded. "Could realistically censor agents" is about intent; this is arithmetic. Censorship 7/15 flagged for October as a downgrade candidate. |
| 2026-09-06 | Data | N/A | WS3b distribution rail: first on-chain read of native WARD. The chain names two of the top four addresses as vesting accounts holding 27.59% of supply between them, and the largest single address is a Hyperlane bridge at 24.98%. Token Distribution Fairness 6/15 held, supported. |
| 2026-09-06 | Data | N/A | WS3b liquidity rail: Bybit spot returns no WARDUSDT (control GRASSUSDT resolves), so it is removed from both venue lists, and the daily-volume-exceeds-market-cap line is restated as turnover near half the cap (52.15% 7-day median). BingX confirmed real at BingX despite being absent from CoinGecko. Liquidity & Access 8/15 held. |
| 2026-09-05 | Data | N/A | Warden was exposed to the critical Cosmos EVM balance-underflow (GHSA-7g4w-cg88-2cq2) and patched it on 23 Aug 2026, five days before the advisory published (commit `754adbe1`, `cosmos/evm` v0.6.2). Six other chains were exploited; no source places Warden among them. No score change. |
| 2026-08-16 | Data | N/A | Correction. The Governance evidence claimed no on-chain proposals; the module holds five, including token holders rejecting a fixed 3% inflation rate. On-chain verdict added; score flagged for the monthly. |
| 2026-08-12 | Data | N/A | WS3 Cosmos on-chain reconciliation (1->3 verdicts). Native supply read on WardenChain (Cosmos-SDK): award denom = 1.052B WARD at 18 decimals, already above the 1B genesis, inflating ~10%/yr per the mint module (established, confirms the uncapped-PoS framing). Open-source confirmed (wardenprotocol Apache-2.0 ~2.2k stars; match). Scores unchanged. |
| 2026-07-07 | Data | N/A | Internal-consistency fix. The community-tax burn rate was stated as 3% in four secondary spots (a con bullet, the deflationary-case line, the Value Accrual score summary, and a key-risk bullet) while the canonical tokenomics and Value Accrual sections said 2%. Corrected all four to 2%, matching the research JSON (`tokenomics.burn_mechanism`). No score change. |
| 2026-06-25 | Data | N/A | DefiLlama-harvest cross-check + staleness refresh. Verified current market data (CoinMarketCap): WARD ~$0.0098, down ~93.9% from its 4 Feb 2026 ATH of $0.1608, mcap ~$2.46M. The review was written ~early March 2026 and called WARD "one month old" with "a 93% crash in one month" throughout; refreshed all instances to reflect that it has now traded ~5 months and the ~94% decline has been SUSTAINED (not a launch dip). Decline figure 93% -> ~94%, IDO loss -> ~98%. The harvest snapshot's market data (collapse from a ~$200M raise FDV) checked out; its newer self-reported usage figures (63.9M agent runs, 15.2M SPEX proofs, $865K revenue) supersede our 2025-report numbers but we still anchor on the independently verified ~35-37M on-chain tx estimate. No score change (caution thesis intact; the sustained decline reinforces it). |