MCP.Market

Warda MCP server

online · 100% uptime
by ArtyKOMarkets·io.github.ArtyKOMarkets/warda·v0.4.2

Economic authority for autonomous agents on Kaspa L1. Never holds a key.

A88/100grade A
What users say
No reviews yet
Be the first
Safety scan
A88/100

full report

Adoption
Growing

0 stars25 downloads/wk

Reviews

Write one

Nobody has reviewed Warda yet.

If you have run it, two minutes of your experience saves the next person an afternoon.

Warda tools (11)

write = sends, deletes, buys or posts
  • warda_build_delegationFree

    Assemble an unsigned transaction that subdivides this grant into a child for a sub-agent, and return the digest to sign with the PARENT's agent key. THIS SERVER NEVER SEES YOUR KEY. A child can only ever be more restrictive than its parent, and the parent reserves exactly what the child receives — authority is subdivided, never created. Narrowing the child's allowlist needs the parent's full member list, because the witness is a path through the parent's tree and a root alone cannot produce one.

  • warda_build_exitFree

    Assemble an unsigned transaction that ENDS this grant and sweeps its balance to the principal. 'revoke' is the emergency stop, signed by the revocation key, effective the moment it confirms — the agent cannot stop it and cannot outrun it beyond the limits it already had. 'reclaim' is the same sweep after the window has closed, signed by the principal key; the covenant enforces the timing with a CLTV, so a reclaim built early simply will not run. THIS SERVER NEVER SEES A KEY. This is the tool a monitor calls when a grant is behaving badly and something has to act.

  • warda_build_settlementFree

    Assemble an unsigned transaction that ends a child grant and returns its UNSPENT remainder to the parent's budget, charging the parent only what the child actually spent. This is the other half of delegation: letting a child expire instead returns the money to the PRINCIPAL — to the human — which is no use to an agent mid-task. Returns TWO digests: the parent's half is signed by the parent's agent key, the child's by the revocation key. THIS SERVER NEVER SEES EITHER. prevRoot is the parent's reserve root from before this child was pushed; warda_build_delegation returns it as parentReserveRootBefore. It cannot be derived from anything on chain — a hash chain pops by preimage.

  • warda_build_spendFree

    Assemble an unsigned Kaspa transaction that spends from this grant, and return the digest to sign. THIS SERVER NEVER SEES YOUR KEY and never signs: it hands back bytes, and you sign the digest wherever your key lives. It also returns the advisory verdict, but it builds the transaction either way — a local rule that is too strict must not be able to block a payment the chain would accept. If the verdict says refused, broadcasting will cost a fee and fail.

  • warda_check_delegationFree

    Check whether a proposed child grant is a legal narrowing of this one. A child may only ever be more restrictive than its parent, and the parent must reserve exactly what the child receives — authority is subdivided, never created. Advisory.

  • warda_check_spendFree

    Check a proposed payment against every rule the covenant enforces, and return the same verdict the chain would. Builds the recipient proof for you. ADVISORY: a 'permitted' answer is not permission — it means the covenant would accept this transaction. A 'refused' answer means no valid transaction exists, so broadcasting one would only waste a fee.

  • warda_find_serviceFree

    Search the Warda service registry for endpoints an agent holding a grant can pay. Filters are a conjunction: a service must offer EVERY capability asked for and be at or under maxPrice. WHAT IS VERIFIED: that the operator controls the key the service is paid at, and that the listing was served from the same host as the endpoint it names. WHAT IS NOT: that the price is real, that the service works, or that it is worth paying — those are the operator's claims about themselves and arrive labelled as such. There is no ranking; do not read the order as one. NETWORK: unlike every other tool on this server, this one makes one outbound request, to the registry and to nothing else. It will not fetch a URL you give it. NOT REQUIRED: discovery is never in the payment path. A listing is a signed manifest at /.well-known/warda-service.json on the operator's own domain — if you already know an endpoint, read that directly and pay it without asking anyone.

  • warda_grant_addressFree

    Derive the on-chain address of a grant from its terms and current state. A grant's address is a HASH of its state, so it MOVES after every spend and every delegation — an address that worked yesterday holds nothing today. Use this to know which address to fund, watch, or look up.

  • warda_grant_authorityFree

    The agent's spending power right now: budget left, epoch headroom, per-transaction cap, and the largest single payment currently permitted. This is what an agent wallet's spending limit would tell you, except that it is not a setting — the covenant enforces these whether or not you ask. Use it before planning a purchase. NOT a chain balance: this server never reads the chain. It reports what the grant you passed says, so a stale descriptor gives a confident wrong number.

  • warda_recover_grantFree

    Read a grant's full state back out of any transaction that spent it, and work out where it went. Use this when a grant's recorded state is stale, wrong, or lost and its address turns up empty. Kaspa's P2SH requires the covenant script to travel in the clear inside every spending transaction, and the grant's state is spliced into that script — so every spend publishes the grant it spent. Pass the transaction as wire JSON, or just the redeem script hex if that is all you have.

  • warda_walletFree

    The whole wallet view in one call: the grant's current address, its spending limits, the largest payment permitted right now, and every address it is allowed to pay. This is the agent-wallet shape — address, limits, spending power — with one difference worth knowing: the limits are not settings this server or your code applies. They are in the script that unlocks the coin, so a payment outside them is not refused, it is a transaction that does not exist. NO CHAIN READ: this server never connects to a node. Every figure comes from the grant descriptor you passed, so an on-chain balance is deliberately absent — reporting one from a stale record is the failure this whole protocol keeps meeting.

Public scan report

scanner v0.1.10 · 2026-10-29 · same rubric, same numbers if you re-run it

no findings
  • Code scan9 source files scanned25/25
  • Live reliabilityremote reachable in 692ms20/20
  • Tool poisoning11 tool descriptions checked15/15
  • Auth qualityopen endpoint, read-only tools10/15
  • Maintenancelast push 33 days ago12/15
  • Maintainer identityregistry namespace matches repository owner6/10
Overall 88/100. Components that don't apply are left out of the denominator. Any critical finding is an F.RubricAppeal a findingJSON

How Warda compares

Against every graded MCP server, from our own scans and uptime checks. Updated after each scan run.

Safety score
88/100
higher than 75% of 48,243 servers
Tools
11
the typical server has 7
Tools that change things
0%
typical: 0%. Fewer write tools means less an agent can break
Uptime
100%
of our recent checks; the typical remote server answers 100%
Response time
604 ms
faster than 26% of 30,027 servers
Downloads a week
25
more than 11% of 12,306 servers with a package

What its tools let an agent do: 11 read

Install Warda in Claude Code, Cursor or VS Code

claude mcp add --transport http warda https://mcp.wardaprotocol.com/mcp
Add to Cursor

What the publisher says

From the Warda repository's README, as published. We do not edit it. Read it on GitHub

Warda

Give agents money. Not unlimited authority.

An open protocol for creating, delegating, verifying and enforcing cryptographic economic grants for autonomous software agents, on Kaspa.

A principal commits funds to a grant and defines what the agent may do with them. The property that matters:

The agent cannot exceed the authority encoded in its grant — even if the agent

itself, its wallet software, or the Warda backend is compromised.

Enforcement is not a policy in a database. It is a Toccata covenant: the settlement layer refuses to produce a valid transaction.

Live on testnet-10

An agent was prompt-injected and told to pay an address outside its allowlist. The network refused the transaction.

Same grant, same address, same key — differing in one field, the payee. Full detail in DEPLOYED.md.

Status: experimental, unaudited, nothing on mainnet

The spend and delegation covenants exist and are proven against TxScriptEngine — the same script engine a Kaspa node uses to validate a transaction. 33 covenant tests, 45 protocol tests, sub-second, no node required.

Nothing here has touched mainnet, and Silverscript itself is pre-v1 and may break without notice.

Quick start

git clone https://github.com/ArtyKOMarkets/warda && cd warda && npm install

npx warda key --out wallet.key                  # fund the address it prints
npx warda wallet                                # what it holds, and what it can fund
echo kaspatest:qq7xj0mpl0p46875mnkzhwatdy478pjkum745srhaey44l9jx566zefjaam3e > payees.txt
WARDA_SK=$(cat wallet.key) npx warda grant --payees payees.txt --budget 10 --max-per-spend 1
npx warda pay https://warda-demo-api.vercel.app/fact

warda grant remembers the manifest, the allowlist and the agent key in .warda/config.json, so everything after it needs a URL and nothing else. Amounts are KAS, not sompi — --budget 1000000000 is a number nobody can check by eye, and getting it wrong by a factor of ten is a grant that is silently ten times too permissive.

A payment to an address that is not in payees.txt does not fail a check. The allowlist is compiled into the script that unlocks the coin at creation, so there is no valid transaction to build — not one the network would reject, none at all. That refusal is the whole product, and the tutorial walks the same commands with the reasons attached.

Checking it, rather than using it

npm run check                              # protocol semantics: typecheck + 45 tests
cd covenant/harness && cargo test          # covenant vs. the node engine: 33 tests
cd covenant/deploy && cargo run -- dry-run # deploy tool, no node needed

What is actually proven

Each row below has a flip test: a spend the engine accepts, with exactly one field changed. Because the baseline passes, the rejection can only be caused by that field.

Shortened. The full README is on GitHub.

Nothing above is checked by us. What we check is on the safety report.

Warda: common questions

Is Warda MCP server safe?
Yes, by our scan: it is graded A (88/100). Read the Warda safety report
How do I install Warda?
It runs remotely at mcp.wardaprotocol.com. Add it to Claude Code, Cursor, VS Code or Claude Desktop with the snippets above, or call it through the mcp.market gateway without installing anything.
Does Warda need an API key?
No secret keys are declared. It reads 1 setting from the environment.
Is Warda maintained?
The last commit was 40 days ago (2026-09-26). The latest release is v0.4.2.
Is Warda up?
100% of our last 28 checks got an answer. We check remote servers about four times a day.
How does Warda compare with other MCP servers?
safety score 88/100 (higher than 75% of 48,243 servers); tools 11 (the typical server has 7); tools that change things 0% (typical: 0%. Fewer write tools means less an agent can break). Full comparison.