Summary
| Item | Detail |
|---|---|
| Applicant | Fibrous, a DEX aggregator that also runs a CoW Protocol solver |
| Project | Fiberscope, open-source analytics for CoW Protocol solvers |
| Today | Live on Base at fiberscope.org, updated every 10 minutes. Code: github.com/Fibrous-Finance/fiberscope (MIT) |
| Already delivered, not charged | The Base indexer and site, methodology, validation against Dune and CoW’s data, and self-hosting documentation |
| This grant funds | The nine other production chains with an all-chains overview; per-solution scores, quote winners and protocol fees; a public JSON API with per-auction data; review and handover, then 12 months of maintenance |
| Type | Hybrid: three milestones over about 5 weeks, then 12 months of maintenance funded by the maintenance stream (vesting COW) |
| Request | 14,000 USDC across three milestones, each paid after sign-off, and 15,000 COW vesting over 12 months |
| Ongoing cost to the DAO after delivery | Only the 15,000 COW maintenance stream (fixed, 12 months). Fibrous pays all servers, RPC and hosting |
| Conflict of interest | Fibrous runs a CoW solver (Base, BNB Chain), disclosed in full below |
Evaluation criteria at a glance
The criteria are those of the 2026 mandate.
| Criterion | Where addressed |
|---|---|
| Strategic Alignment (“solver competition”) | Fiberscope measures solver competition (entries, wins, efficiency and surplus) on Base today, and on every CoW chain with this grant. See Problem and Goals |
| Measurable Impact | Coverage, accuracy and freshness targets, with quarterly adoption reporting. See Measurable outcomes |
| Feasibility | Base is live and validated. The grant extends a running system and requires no protocol changes. See Accuracy and Technical approach |
| Technical Expertise | Fibrous runs a CoW solver and a DEX aggregator. Fiberscope agrees with CoW’s Solver Info dashboard to within about 0.1% and credits 24,534 of 24,534 batches to the winner in CoW’s competition data. See Experiences and qualifications and Accuracy |
Author(s)
- Kermo (forum: @kermo, GitHub:
0xKermo), Fibrous: main point of contact - Enes (GitHub:
ahmetenesdur): full-stack engineer at Fibrous, working across its aggregator frontend, backend services and on-chain data infrastructure. Designed and built Fiberscope end to end, from the settlement indexer and validation pipeline to the site, and leads its development under this grant. - Team: Fibrous (
fibrous.finance, GitHubFibrous-Finance, X@FibrousFinance)
Experiences and qualifications
- Fibrous is a DEX aggregator live on Starknet, Base, HyperEVM, Citrea and Monad, with more than +500M USD in volume, +900K users and 2.1M trades (
fibrous.finance, October 2026). It runs its own routing engine, chain indexers and transaction simulator in production. - CoW Protocol solver. Fibrous is listed as active in CoW’s solver registry on Base, where it has settled batches since July 2026, and on BNB Chain.
- Fiberscope is the work this grant extends. It agrees with CoW’s Solver Info dashboard to within about 0.1% and credits every batch checked to the winner in CoW’s competition data; see Accuracy.
Grant Description
Category
Solver competition transparency: analytics and a public API for every chain CoW Protocol serves.
Background
Fibrous began solving on Base in July 2026. In September 2026, the Base view of CoW’s Solver Info dashboard on Dune stopped showing new settlements, and from late September free Dune accounts became view-only. In early October 2026, we built Fiberscope as a public, open-source replacement for Solver Info’s Base view. We validated Fiberscope against Dune and CoW’s competition data.
Problem
We made these observations while logged out on 5 October 2026 at 09:50 UTC.
- Solver Info is only as fresh as its last paid run. Solver Info has long been the public reference for solver competition. A Dune dashboard shows cached results until someone with a paid Dune plan re-runs its queries, and this applies to every chain alike. At the time of our observation, the default Ethereum view had last been run on 3 October 2026. On the other chains, the newest settlement in its “V3: Active Solvers” table was:
-
- 14 September 2026 on Base (Base view), Avalanche, BNB Chain, Gnosis Chain, Ink, Linea, Plasma and Polygon;
- 20 September 2026 on Arbitrum One.
On Avalanche, Gnosis Chain and Linea, some panels return “Dune.cowprotocol.solvers does not exist or it is private”.
- Most viewers cannot refresh Solver Info. Since September 2026, free Dune accounts are view-only after a 14-day trial; running queries and using Dune’s API require a paid plan (Dune changelog, August 2026 entry). How current any chain’s view is therefore depends on whether, and when, someone with a paid plan last ran it.
- We found no public source of entry and win rates besides Fiberscope. Dune’s Spellbook has CoW batch and trade tables for every chain but no solver-competition tables, and, as far as we can find, CoW’s auction-level uploads to Dune list winners only. Losing solutions are available only through CoW’s solver-competition API, one auction per request. Outside Fiberscope’s Base view, we found no public view of how often each solver enters auctions and how often it wins when it enters.
Dune’s tables are accurate; Fiberscope is validated against them. The gaps are freshness, free access and auction-level data. Re-running Solver Info on a paid plan, or scheduling it, would restore freshness. It would not provide entry rate and win rate, which require losing solutions that are not on Dune.
The following tools remain useful; Fiberscope does not duplicate them:
- Solver Info itself, whenever it has been run recently.
- CoW’s Solver Rewards by Week, Solver and Chain, which covers rewards.
- CoW Explorer explains single orders and settlements and lists solvers in its solver directory. Fiberscope links to CoW Explorer for every settlement it lists.
What already exists (built and self-funded, not charged)
Fiberscope for Base is live at fiberscope.org and updated every 10 minutes. It keeps 180 days of settlements and 90 days of auctions and surplus. In a longer window, figures that need auction or surplus data cover only the last 90 days, and the site says so.
| View | What it shows |
|---|---|
| Who is winning | Each solver’s share of batches, trades or volume over 24 hours or 7, 30, 90 or 180 days, with rank changes, entry rate and win rate; per-solver detail view; CSV or Markdown export |
| Is it changing | Daily totals across all solvers, the leading solvers’ daily share, and how long the current leader has led |
| Who enters, who wins | Entry rate and win rate per solver over auctions that ended in a settlement; average solvers and solutions per auction; the latest 48 auctions, linked to CoW Explorer |
| Latest settlements | Pair, trades, DEX swaps, volume, surplus, transaction cost and gas for the 50 latest batches, linked to CoW Explorer and Basescan |
| How efficiently | Gas, transaction cost (including the L1 data fee) and DEX swaps per trade for each solver, compared with the network average |
| Trader surplus | Each solver’s surplus in basis points of volume. Surplus above 10% of a trade’s value (CoW’s own cutoff) comes from limits set far from the market and is shown separately |
| Solver registry | CoW’s full Base solver registry (43 entries, 22 active), with addresses and last settlement |
| Methodology | Definitions for every figure on the page and in docs/methodology.md in the repository |
Example (Base, 5 October 2026): over 7 days, an average of 8.3 solvers entered each settled auction, submitting about 10 solutions between them.
All data sources are public:
GPv2Settlementevents, receipts and calldata, read over RPC;- CoW’s API,
GET /api/v2/solver_competition/by_tx_hash/{tx}; - CoW’s solver registry;
GPv2AllowListAuthentication.isSolver;- Chainlink ETH/USD;
- ERC-20
symbol().
Architecture: an indexer (Node 24, SQLite, Docker) writes one JSON file per network to Cloudflare R2. A Next.js site on Cloudflare Workers renders it.
Accuracy
From docs/methodology.md#validation in the repository:
| Check | Result |
|---|---|
| Against Solver Info’s Base figures, blocks 50,409,187–51,705,186 (the 30 days to 23 September 2026) | Batches 76,954 vs 76,873 (+0.11%); trades 80,892 vs 80,808 (+0.10%); DEX swaps 116,166 vs 116,034 (+0.11%); gas per trade 730.7K vs 730.6K (+0.01%) |
| Attribution against CoW’s competition data, 26 September–3 October 2026 | 24,534 of 24,534 batches credited to the winning solution’s solver, including all 1,230 sent through CoW’s flash-loan router |
| Order terms behind surplus, on a sample of 407 trades covering every calldata route to the settlement contract (direct calls, CoW’s flash-loan router, solver contracts) | All 407 decoded trades reproduce their Trade events; 80 of 80 orders checked agree with CoW’s order API |
Solver Info caches panels separately: the Base figures cover the 30 days to 23 September 2026, while its Active Solvers table for Base stops at 14 September. Over the same blocks, Fiberscope counts about 0.1% more batches, trades and DEX swaps. Before M1’s review, at our own cost, we will recheck Base against the Spellbook tables used to measure the new chains (see M1) and publish an explanation of this difference.
How Fiberscope relates to existing tools
| Capability | Solver Info (Dune) | Solver Rewards (Dune) | CoW Explorer | Fiberscope |
|---|---|---|---|---|
| Current batches, trades, volume and share per solver | Yes, as of its last paid run | No | No | Yes, every 10 minutes: Base now, all 10 chains after M1 |
| Entry rate and win rate per solver | No | No | No | Yes |
| Gas and DEX swaps per trade per solver | Yes, as of its last paid run | No | No | Yes, with transaction cost including the L1 data fee, against the network average |
| Surplus per solver | Total surplus, as of its last paid run | No | No | Yes, with surplus from limits far from the market shown separately |
| Single orders and settlements | No | No | Yes | Links to CoW Explorer |
| Solver rewards | Yes | Yes | No | No (out of scope) |
| Free API for these figures | No (Dune’s API needs a paid plan) | No (same) | Not applicable | Yes, after M2 |
Goals of this grant
- Every production chain. Extend Fiberscope from Base to the nine other chains CoW Protocol’s API serves in production: Ethereum, Gnosis Chain, Arbitrum One, Avalanche, Polygon, BNB Chain, Linea, Plasma and Ink, with an all-chains overview. These are the mainnets marked “(Prod)” in the orderbook API’s OpenAPI specification; Sepolia, a testnet, is out of scope.
- Public API. A documented, versioned JSON API with the same figures as the site, free to use without a key.
- Reliability. Run and maintain the site and the API for 12 months after delivery, with infrastructure paid by Fibrous.
The code is already open source under MIT; all grant work stays under MIT.
Technical approach
We replace the indexer’s Base-specific assumptions for each new chain. Each new chain passes validation before launch.
| Base today | Each chain after this grant |
|---|---|
| Timestamps computed from block numbers (Base’s 2-second block time is exact) | Timestamps from block headers where block time varies |
| Data up to Base’s safe block | Data up to each chain’s newest block safe from reorganization: its safe or finalized tag, or a documented confirmation depth where those tags lag too far behind for 10-minute updates |
| Cost: gas used × effective gas price, plus the OP Stack L1 data fee | Each chain’s fee model: the L1 data fee on Ink; on Arbitrum One, the L1 share of gas used is left out of gas per trade but kept in cost; gas used × effective gas price elsewhere |
| USD from Chainlink ETH/USD on Base | Chainlink’s USD feed for the native token (ETH, xDAI, AVAX, POL, BNB, XPL) where one exists on the chain; otherwise the USD rate implied by CoW’s own auction prices for a USD stablecoin |
| Base contract addresses | Settlement, allow-list and flash-loan router addresses for each chain |
| Base hard-coded in the site’s links and labels | Every view per network, with the network in shareable URLs and exports |
We repeat the Base validation checks for each chain. We compare totals with Dune’s Spellbook tables (cow_protocol_<chain>.batches and .trades, which exist for all ten chains) and check attribution against CoW’s competition data.
Impact
- Solver competition. A current, free view of who competes, who wins and at what cost on every CoW chain. Today, how current Solver Info is on any chain depends on when someone with a paid Dune plan last ran it, and we found no other public source of entry and win rates.
- Governance. Delegates and the core team can follow solver concentration on each chain; on 5 October 2026, for example, one solver settled 38.5% of Base batches over 24 hours. This informs discussions about solver rewards and the solver set.
- Prospective solvers. A team choosing a launch chain can see how many solvers enter, how concentrated the wins are and how efficient incumbents are on each chain.
- Builders and researchers. A free API instead of a paid Dune plan.
Measurable outcomes
We report outcomes quarterly in this thread during maintenance.
| Measure | Target |
|---|---|
| Chains covered | 10 of 10 production chains |
| Accuracy, per chain | Batches and trades within 0.5% of Dune’s Spellbook over a 30-day block range; a larger difference requires a documented cause and the reviewer’s approval. At least 99.9% of batches credited to the solver named as the winner in CoW’s competition data |
| Freshness, per chain | Data at most 30 minutes old at least 99% of the time, measured by M2 monitoring |
| Accuracy reports | Acknowledged within 3 business days; fixed or explained within 14 days |
| Adoption | Monthly visits and API requests (Cloudflare analytics, known bots excluded), distinct API clients, and links from forum threads. Reported without a target |
Alignment with Grants DAO values
| Value | How this application meets it |
|---|---|
| Open source | MIT since launch; all grant work under MIT |
| Milestones | Three milestones, each with acceptance criteria a reviewer can check |
| Price transparency | A fixed price for each milestone and for the maintenance stream; no upfront payment |
| Sustainability | Fibrous operates the service and pays for its infrastructure; 12 months of maintenance through the COW maintenance stream; Docker and documentation let anyone redeploy it |
| Simplicity | Extends a working system; no protocol changes or new infrastructure for the DAO |
| Documentation | Methodology, self-hosting and API documentation and an OpenAPI specification |
| Flexibility | Any change to scope or price goes to the Grants Committee |
Conflict of interest disclosure
- The conflict. Fibrous runs a CoW Protocol solver and competes in the auctions Fiberscope measures. In CoW’s solver registry, Fibrous is active on Base (settling since July 2026) and on BNB Chain (registered on 31 August 2026); its Arbitrum One entries are inactive.
- Benefit to Fibrous. The footer credit and the site’s visibility. No data advantage.
- Data. Fiberscope uses only the public data sources listed under What already exists. It has no access to non-public CoW data. Fiberscope gives Fibrous nothing another solver cannot compute from the same sources.
- Equal treatment. Every solver is ranked by the selected measure alone. Fibrous gets no special position or color. Apart from its data rows, like any solver, Fibrous appears on the site only in the footer credit: “Built by Fibrous, which also runs a solver.” We do not feature Fibrous in examples or screenshots.
- Checkable. The code and methodology are public. All listed settlements link to CoW Explorer and the chain’s block explorer, and
node src/cli.ts window --from <block> --to <block>recounts any block range from the database. - Changes. Methodology changes are made in public pull requests. We invite other solver teams and CoW’s core team to review the methodology in this thread.
- Exit. At the Grants Committee’s request, we transfer the repository and domain to a CoW-controlled organization.
Ownership and sustainability
- Fiberscope runs on Fibrous’s AWS and Cloudflare accounts. Fibrous pays for its infrastructure: servers, RPC endpoints, Cloudflare, and any Dune usage for validation. The grant adds no recurring DAO infrastructure cost.
- Fiberscope currently runs on Base, and its running costs are modest (
docs/development.md#running-costsin the repository). As the grant extends it to all ten chains, Fibrous will continue to cover all running costs, including servers, RPC plans and Cloudflare. - The code stays MIT-licensed, as the Grant Agreement allows. Ownership of grant work follows the Agreement’s IP terms.
- Fiberscope is an independent project, not affiliated with CoW DAO.
- After the maintenance stream ends, Fibrous intends to keep operating Fiberscope at its own cost. Before any shutdown, we will give at least 60 days’ notice in this thread and offer a handover.
Risks and mitigations
| Risk | Mitigation |
|---|---|
CoW’s API changes (for example, /api/v1/solver_competition/latest now returns 404) |
Fiberscope uses /api/v2/solver_competition/by_tx_hash. Monitoring from M2 alerts on stale data, and maintenance covers adapting within a week |
| RPC limits or outages | Paced, retrying clients and paid endpoints. The site labels data “Delayed” once the newest block is more than 30 minutes old, so stale figures are never shown as current |
| CoW API rate limits while new chains load their history | A paced client (5 requests a second today) loads history between the regular updates, keeping live chains current while loading a new chain’s history |
| Chain-specific fees and finality | A chain goes live only after passing validation |
| No Chainlink feed for a native token on a newer chain | USD rates from CoW’s own auction prices for a USD stablecoin, as described in Technical approach |
| Neutrality concerns | The disclosure above, public code and methodology, and the transfer offer |
| A single team | Docker, documentation and a self-hosting guide let another team take over |
Out of scope
- Solver rewards and payouts, which CoW’s Solver Rewards dashboard covers.
- Bonding pools, testnets (Sepolia), and chains the API no longer serves (Lens).
- Any non-public data.
Type of Grant
Hybrid: milestone-based development (M1–M3, about 5 weeks), followed by 12 months of maintenance funded by the COW maintenance stream, which starts when M3 is completed, as in the Dune Governance Dashboard grant (forum topic 2648).
Milestones
Delivered before this application and not charged: Fiberscope for Base, with the live site, open-source code, methodology and validation.
| Milestone | Title | Due date (*) | Funding request |
|---|---|---|---|
| M1 | Multi-chain indexer: all nine new chains, an all-chains overview and per-solution scores | Week 3 | 9,000 USDC |
| M2 | Public JSON API v1 with per-auction data, protocol fees, and monitoring | Week 4 | 2,500 USDC |
| M3 | Review, fixes and handover | Week 5 | 2,500 USDC, and the maintenance stream: 15,000 COW vesting over 12 months |
(*) Weeks after the proposal passes on Snapshot. Each milestone has a fixed price for its deliverables and is paid after the reviewer’s sign-off. There is no upfront payment.
Specifics: M1, multi-chain indexer for all nine new chains
3 weeks, 9,000 USDC.
Deliverables:
- The indexer configured per network, as described in Technical approach.
- BNB Chain first, then Ethereum, Gnosis Chain, Arbitrum One, Avalanche, Polygon, Linea, Plasma and Ink, each indexed with 180 days of settlements and 90 days of auctions and surplus, updated every 10 minutes.
- Per-solution data stored for every auction: solver, ranking, score, reference score, filtered-out status and winner.
- Quote winner per order from the order’s quote, compared with the solver that settled it; quote-winner figures per solver on the site.
- A network selector on the site, every view per network, and the network in shareable URLs and exports.
- An all-chains overview: totals per chain and each solver’s activity across chains.
- A validation section for each new chain in docs/methodology.md.
Acceptance:
- All nine chains are live on the public site and update every 10 minutes.
- For each chain, over a 30-day block range: batches and trades are within 0.5% of Dune’s Spellbook, or a larger difference has a documented cause the reviewer accepts; and at least 99.9% of batches are credited to the solver named as the winner in CoW’s competition data. Results and queries used are published.
- On a published sample of orders on each chain, quote winners match CoW’s order API.
- The overview’s totals equal the sum of the per-chain figures.
Specifics: M2, public API v1 and monitoring
1 week, 2,500 USDC.
Deliverables:
- Versioned JSON endpoints under
/api/v1/for every live network, covering:- per-solver figures for each window (24h, 7d, 30d, 90d, 180d) and measure (batches, trades, volume): share and rank, entry rate and win rate, gas, transaction cost and DEX swaps per trade, surplus, and quote wins;
- per-auction rows: every solution with its solver, ranking, score, reference score, filtered-out status and winner;
- protocol and partner fees per settlement;
- daily series;
- the latest auctions and settlements;
- the solver registry.
- Protocol fees per settlement from `executedProtocolFees` on CoW’s trades API, broken down by fee policy, with partner fees separated using each order’s app data.
- An OpenAPI 3.1 specification and documentation with curl and JavaScript examples.
- Edge caching, rate limiting and CORS. No API key.
- Monitoring that alerts when a network’s data is more than 30 minutes old.
Acceptance:
- Every endpoint responds for every live network; responses validate against the specification.
- On a published sample of settlements on each chain, protocol and partner fees match CoW’s trades API.
- The API and site show the same figures for the same window and read the same data.
Specifics: M3, review and handover
1 week, 2,500 USDC. The maintenance stream starts on completion.
Deliverables:
- A feedback round in this thread with CoW’s core team and solver teams, and the agreed fixes that fit within M3’s week. Larger requests become follow-up proposals.
- A final review of the methodology, self-hosting and API documentation.
- A final report in this thread.
- The repository and domain transferred to a CoW organization, if the Grants Committee requests it.
Acceptance: the fixes agreed in the feedback round are merged, the final report is posted, the methodology, self-hosting and API documentation are complete, and the assigned reviewer signs off.
Maintenance (12 months after M3, funded by the COW maintenance stream)
- Keep the site and the API running for every delivered chain, updated every 10 minutes, to the targets in Measurable outcomes.
- Add up to two chains that CoW launches in production during the year. Each goes live within 4 weeks of the later of its CoW launch and the availability of an RPC endpoint and a USD price source. If CoW launches more than two, the Grants Committee chooses which. More by agreement.
- Adapt to breaking changes in CoW’s API within one week.
- Post a quarterly update in this thread covering freshness, per-chain validation, adoption and running costs.
- If a quarterly update shows a freshness or accuracy target missed on any chain, and that chain does not meet it again within 30 days of the update, the Grants Committee may stop the maintenance stream. Unvested COW then stays with the Grants Treasury.
Length
- Commencement: when the proposal passes on Snapshot.
- Completion: M3, 5 weeks after commencement.
- Maintenance: 12 months after M3. We request a length exception for this period. Development completes well within the 6-month limit, and maintenance is funded by the COW maintenance stream, as in the Dune Governance Dashboard grant (forum topic 2648).
- The maintenance stream would run past the Grants Committee’s current mandate, which ends on 31 January 2027. We ask the committee to confirm that it can set up a stream of this length.
Funding Request
- 14,000 USDC, paid in USDC on Ethereum mainnet to the address below, each milestone after the reviewer’s sign-off, in the Grants Committee’s monthly payout cycle. No upfront payment is requested.
- 15,000 COW for 12 months of maintenance. The Grants Committee sets up the maintenance stream with linear vesting from the Grants Treasury on Ethereum mainnet (for example through LlamaPay) when M3 is completed. The Grants Committee can stop the maintenance stream as described under Maintenance. At 0.158 USD per COW (CoinGecko, 5 October 2026), this is about 2,400 USD.
| Item | When | Amount |
|---|---|---|
| M1: multi-chain indexer, nine chains and the overview | Week 3 | 9,000 USDC |
| M2: public API and monitoring | Week 4 | 2,500 USDC |
| M3: review and handover | Week 5 | 2,500 USDC |
| Total | 14,000 USDC | |
| Maintenance, 12 months after M3 | 15,000 COW, vesting | |
| Base build, already delivered | Not charged | |
| Infrastructure: servers, RPC, Cloudflare, Dune for validation | Paid by Fibrous |
Rationale: prices are fixed per deliverable. For comparison, the Dune Governance Dashboard grant (forum topic 2648) was 10,000 xDAI and 10,000 COW, with the COW vested over 12 months, for four one-week milestones. This grant covers ten chains, a public API and a year of operation, with infrastructure paid by Fibrous.
Ethereum Mainnet Address (to receive the grant)
To be provided before the proposal moves to Snapshot (Ethereum mainnet, Fibrous). The template asks for a Gnosis Chain address, but under CIP-82 the Grants Treasury pays from Ethereum mainnet. If the address is a Safe, it is deployed on Ethereum. Recipient: Fibrous.
Other Information
- Live site:
fiberscope.org - Code:
github.com/Fibrous-Finance/fiberscope - Methodology and validation:
docs/methodology.mdin the repository - Self-hosting and running costs:
docs/development.mdin the repository - Solver Info, Base view:
dune.com/cowprotocol/solver-info?blockchain_edb9f7=base - We can demonstrate Fiberscope on a call.
Terms and Conditions
By submitting this grant application, I acknowledge and agree to be bound by the CoW DAO Participation Agreement and the CoW DAO Grant Agreement Terms.