Application: Polkagent

[aurexav]

Project Abstract

• Tag line (AI focus): Turn Substrate metadata into a few AI‑ready macro tools and a multi‑turn Planner, so agents and humans can query chain data safely in natural language.

• Brief description:
  Polkagent exposes a small, stable macro‑tool surface via a Rust CLI + HTTP gateway, plus an AI Planner for multi‑turn task solving. Instead of thousands of fine‑grained endpoints, we provide four high‑level tools designed for LLM/agent integration and manual CLI use.

Grant level

• Level 1: Up to $10,000, 2 approvals
• Level 2: Up to $30,000, 3 approvals
• Level 3: Unlimited, 5 approvals (for >$100k: Web3 Foundation Council approval)

Application Checklist

• The application template has been copied and aptly renamed (project_name.md).
• I have read the application guidelines.
• Payment details have been provided (Polkadot AssetHub (USDC & DOT) address in the application and bank details via email, if applicable).
• I understand that an agreed upon percentage of each milestone will be paid in vested DOT, to the Polkadot address listed in the application.
• I am aware that, in order to receive a grant, I (and the entity I represent) have to successfully complete a KYC/KYB check.
• The software delivered for this grant will be released under an open-source license specified in the application.
• The initial PR contains only one commit (squash and force-push if needed).
• The grant will only be announced once the first milestone has been accepted (see the announcement guidelines).
• I prefer the discussion of this application to take place in a private Element/Matrix channel. My username is: @_______:matrix.org (change the homeserver if you use a different one)

[diogo-w3f]

@aurexav thanks for applying. Could you please change your application to comply with our template? Here are the changes needed:

• Default milestone deliverables must follow the template. Each table should keep 0c. Testing and Testing Guide (explain coverage, how to run tests) and 0d. Docker (artifacts to run everything). Currently 0c is “Methodology” and 0d is “Infrastructure” in all milestones.
• Lines starting with > are discouraged per template instructions; remove both > ❗ callouts to stay compliant.
• Payment block should mirror the template wording (- Payment Details: with nested - DOT: … and - Payment: …). Right now it’s under * Payment Address and could be flagged as off-format.

Furthermore, can you take a read at our new guidelines and help us to outline how your application fits on it? How your application can help to increase the usage of DOT token?

[aurexav]

Thanks for your quick reply, @diogo-w3f.

@aurexav thanks for applying. Could you please change your application to comply with our template?

I apologise for not noticing that. I used AI to improve my English grammar.

Fixed.

Furthermore, can you take a read at our new guidelines and help us to outline how your application fits on it? How your application can help to increase the usage of DOT token?

Of course, I already check that. I also add more detail to elaborate on the idea of this project.

More specifically, https://github.com/w3f/Grants-Program/pull/2686/files#diff-81b9248ff1c8c625aa5d7a3d66d0855663d97818dfab1449a4c7eccd56582beaR133-R164.

[diogo-w3f]

@aurexav Thanks for the improvements. I’ll mark the application as ready for review. The committee will take a look, and new questions may arise.

[Noc2]

Do you already have anyone who wants to use the tool or has specifically requested it? What's your target persona? Your current target audience is not very concrete.

[aurexav]

Do you already have anyone who wants to use the tool or has specifically requested it? What's your target persona? Your current target audience is not very concrete.

I already have clear and broad target users, and the motivation for this project comes directly from real, day-to-day needs we face inside our own development workflow.

---

1. People who will use it right now (existing, real demand)

• Node / runtime / DevOps developers (including my own team)
  This is a daily pain point: constantly typing long RPC curl commands, remembering parameters, copying docs every time.
  For example, we often need to run something like:

  curl https://docs-demo.dot-mainnet.quiknode.pro/ \
   -X POST \
   -H "Content-Type: application/json" \
   --data '{"method":"system_health","params":[],"id":1,"jsonrpc":"2.0"}'

  Even while writing this documentation, I sometimes have to look up the exact command again.
  With Polkagent, I could just say:
  “Check the system health of this endpoint,” and get a structured response immediately.
  That’s exactly what Phase 1 already delivers — a CLI + Planner that performs read-only Substrate RPC queries (e.g., system.health, system.account) safely and reproducibly.

• dApp, wallet, and explorer engineers / support teams
  They frequently need quick checks for usable balances (free − locked − reserved), identify which pallet a hash might belong to, or list storage prefixes on a chain.
  All of these are covered in Phase 1 via natural-language, read-only commands.

• Governance & community operators, and newcomers
  They often ask “How much can I use right now?” or “Which pallet produced this hash?” — tasks that currently require understanding storage or event internals.
  Polkagent answers these naturally, without needing that background.

---

2. Clearly defined target personas (by priority)

1. Substrate node/runtime/DevOps engineers — primary focus, heavy daily RPC usage
2. dApp/wallet/explorer engineers and support teams — need reproducible queries and audit logs
3. Governance & community operators — require quick answers to balance and period questions
4. Newcomers and advanced users — want chain data without installing or learning CLI tools

These are not vague personas; each has clear jobs-to-be-done that Phase 1 (read-only) already supports.

---

3. Market signals and prior proposals

Before designing Polkagent, I studied other relevant proposals:

• [w3f/Grants-Program #2535](MCP server for Substrate-based chains #2535) — a dev-kit aimed at developers but still quite technical to use.
• [w3f/Grants-Program #2549](Add application from Elastic Labs for Polkadot Agent Kit #2549) — had a broader audience but was discontinued.

Together they show two facts:
(1) the need clearly exists, and
(2) there is still no unified, easy interface.

Polkagent fills that gap with a macro-tool + Planner + natural-language design, starting with read-only capability and common queries first.

---

4. More advanced use cases (future phases with indexer integration)

• Ask “Where did the last 100 DOT in my account come from?” — complete fund tracing.
• Query “Show me the cross-chain XCM transfer path for this asset,” instead of manually checking event APIs across chains.
• Developer debugging: “Within this block range, which GRANDPA validators missed their vote/comission submission?”
• Governance automation: commenting in GitHub issues like
  /polkagent what's current referenda period of xxx?
  to get real-time data responses.

These are indexer-backed, multi-data-source conversations, which will be introduced later — but they rely on the same macro-tool + Planner framework built in Phase 1.
The path is clear and incremental.

---

In the AI era, a tool like this is essential: it removes friction, attracts new users, keeps existing builders productive, and ultimately strengthens the whole Polkadot ecosystem.

[aurexav]

Hi, any feedback?

[keeganquigley]

Hi @aurexav sorry for the long delay here. I see that Subalfred was archived. What happened there and how would we be able to ensure that it doesn't happen with this project as well? I think adoption would likely be an even more important factor than the tech.

[aurexav]

Hi @aurexav sorry for the long delay here. I see that Subalfred was archived. What happened there and how would we be able to ensure that it doesn't happen with this project as well? I think adoption would likely be an even more important factor than the tech.

I totally get what you mean.

This project is essentially the successor to Subalfred in the AI era — in some sense you can think of it as Subalfred V2. It will include basically all of Subalfred’s core capabilities (with a few exceptions, e.g. exporting chain state — polkadot-sdk now supports that natively).

Since the primary focus here is AI-driven functionality, I’m planning to do a rebrand, and I think “Polkagent” is a better name for what this project is becoming. That said, I’m also considering unarchiving Subalfred and renaming the repository instead.

---

Since we last update of this is two months ago, the AI landscape has moved quickly—and I’ve done a lot of research in the meantime, so I may have some new ideas to propose.

And of course, I’ll take your concern into account. I’ll polish the proposal this week and make everything clearer.

---

My apologies! I’m quite busy this week and I need to postpone this for another week.

[semuelle]

Closing due to inactivity.