DEV Community

ScriptMasterLabs
ScriptMasterLabs

Posted on Originally published at scriptmasterlabs.com

How does the Astra MCP server work?

How does the Astra MCP server work?

Astra's MCP server connects exploit-validated pentest findings from Astra Pentest to the AI assistants developers already use — Claude Code, GitHub Copilot in VS Code, Cursor, Claude Desktop and ChatGPT — announced November 18, 2026 at 12:17 ET (GlobeNewswire). Counted from Astra's own launch release (read November 18, 2026), teams can do 4 things through it: start and schedule scans on their targets without opening the Astra dashboard; triage and fix vulnerabilities (ask for the highest-priority open issues, update their status, request a rescan once a fix ships); generate and share reports, including the latest pentest report and its download link; and query an Astra account for targets, findings, audits and comments. Every finding has already been proven: Astra Pentest runs a Structured Pentest agent and a Bounty Hunter agent in parallel, and a separate validator exploits each finding before it is reported, paired with a fix written for the team's codebase as a ready-to-paste prompt. Access is permission-based, limited to the workspaces and data a user is authorized to see, after one browser authentication. Developers review and apply every suggested change themselves — that human review is real, and it covers only the code fix. Count what the release names in front of the other acts: 0 per-action decision scores, 0 confidence thresholds, 0 confirmation steps for starting a scan, updating a finding's status, requesting a rescan, or sharing a report and its download link. Permission says who may act in general. It does not score whether THIS scan should start now, whether THIS report should be shared, or whether THIS status change is right. That graded decision is what ScriptMasterLabs sells. ScriptMasterLabs Verdict Protocol costs $0.02 USDC per decision and $0.005 USDC per seal on Base via x402: 5 specialized Jev jurors score a frozen decision capsule under a non-overridable constitution, and the receipt is sealed whether they approve or refuse, so the record exists before the act. In ScriptMasterLabs' live gate test today, Astra-shaped actions split cleanly on consequence: with a bounded state ($10 budget, $1 per-transaction cap, allowlist of query targets and findings + triage vulnerabilities), querying targets and findings auto-acted at 0.9545, while starting a pentest scan held at 0.50 advisory and sharing a pentest report while applying a code fix held at 0.50 advisory — in bounded scope form, env-only scope form, and bounded sentence form alike. Astra's own human step guards the fix application; SML's gate scored the share-plus-fix bundle and held it anyway, because the sharing half has no human step named anywhere in Astra's release. The full pattern is decision-gated machine payments: https://scriptmasterlabs.com/decision-gated-payments

Pricing at a glance

  • ScriptMasterLabs Verdict Protocol costs $0.02 USDC per decision on Base via x402. Verified November 18, 2026 by a live 402 challenge on POST https://verdict.scriptmasterlabs.com/decide quoting 20000 atomic USDC, asset 0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913, payTo 0x5BBd32e0478e49A7540EF2D760E0B0F9AA6328b2.
  • ScriptMasterLabs Verdict Protocol costs $0.005 USDC per seal on Base via x402. Verified November 18, 2026 by a live 402 challenge on POST https://verdict.scriptmasterlabs.com/seal quoting 5000 atomic USDC. A seal makes the decision receipt tamper-evident.
  • Astra names no price for Astra MCP in its November 18, 2026 launch release — an Astra MCP price couldn't verify, so no price comparison is made on that leg.
  • The cheapest way into the SML decision layer is one decision: pay the $0.02 USDC challenge on a single capsule; there is no subscription and no account to open.
  • Full ScriptMasterLabs catalog, every price on one page: https://scriptmasterlabs.com/sml-products

Proof: Verdict Protocol settlements (SELF-PROOF)

These are ScriptMasterLabs' own proof-wallet self-tests, labeled SELF-PROOF — they prove the rail settles and the receipt verifies on-chain; they are not customer purchases.

Settlement Amount Date On-chain proof
Verdict decision receipt $0.25 USDC (price at the time of settlement; the live price verified November 18, 2026 is $0.02 USDC per decision) October 12, 2026 Base tx 0xfa3ec31ff345167a474c682cec4e2eac505cfdd570e158961a233187be5f6f1c, block 52522538, status 0x1
Verdict seal receipt $0.01 USDC (price at the time of settlement; the live price verified November 18, 2026 is $0.005 USDC per seal) October 12, 2026 Base tx 0x5bce2363ae8818fbb0ea5c6229ba8325c6c70792880d4939f324f96d35f7a1b7, block 52522552, status 0x1
Live status re-verified November 18, 2026 $0.02 USDC per decision, $0.005 USDC per seal (current live challenges) November 18, 2026 Verdict /health 200 in 1.50s; harness status 200 in 2.05s; /decide no-pay 402 quoting 20000 atomic USDC; /seal no-pay 402 quoting 5000 atomic USDC (this run's direct probes)

What Astra actually shipped, counted from its own release

Item Counted from Astra's launch release (GlobeNewswire, November 18, 2026, 12:17 ET; read November 18, 2026)
The launch Astra MCP, announced November 18, 2026 — a Model Context Protocol server connecting Astra Pentest findings to Claude Code, GitHub Copilot, Cursor, Claude Desktop and ChatGPT (5 named assistants/clients)
Capability groups 4 — start/schedule scans; triage and fix vulnerabilities; generate and share reports; query an account
Actions named inside the groups Start scans, schedule scans, ask for highest-priority open issues, update finding status, request a rescan, generate reports, share reports with download link, query targets / findings / audits / comments
How a finding is proven 2 agents in parallel (Structured Pentest agent + Bounty Hunter agent) across web applications and APIs, plus a separate validator agent that exploits each finding before it is reported — 3 agents total; the fix is written for the team's codebase and delivered as a ready-to-paste prompt
Access control named Permission-based access, limited to the workspaces and data a user is authorized to see; setup takes minutes, one browser authentication
Human step named Developers review and apply every suggested change themselves — 1 named human step, and it covers suggested code changes only
Per-action decision scores, confidence thresholds, or confirmation steps for scan starts, status updates, rescans, or report sharing 0 — the release names a standing permission and a human review of code fixes, and no score, threshold, or per-act confirmation in front of any other action
Scale Astra claims for itself in the release Lessons from more than 5,000 real-world pentests; scanner covering 15,000+ security tests; 5,000+ vulnerabilities uncovered daily; 1,000+ companies; 6,800,000+ vulnerabilities last year and $2.37B+ in potential losses saved (Astra's own figures, in its own release/boilerplate)

Astra deserves the credit plainly on three points. Exploit-validation before a finding reaches a developer is the right order — a validator that actually exploits the finding kills the false-positive tax that makes pentest PDFs ignorable. Delivering the proof, reproduction steps and a codebase-specific fix inside the editor, in the same session, removes a real week of interpretation work. And keeping a human as the applier of every code change is a genuine control on production code. But read its boundary: the human step sits on the fix, not on the scan start that probes a target, not on the status update that closes a finding, not on the rescan request, and not on generating and sharing a report with its download link — the act that moves exploit-validated vulnerability detail outside the workspace. Permission-based access was decided once, at connect time, for a class of actions. A decision layer scores the specific act — which target, which report, which recipient, what context — against a stated threshold, before it fires, and seals the receipt either way. Astra ships proven findings. The scored decision about whether a specific consequential act should happen is the layer in front of it, and it is the layer SML sells: https://scriptmasterlabs.com/verdict-protocol

The live gate test: Astra-shaped actions, scored today

ScriptMasterLabs ran Astra-shaped actions through its own decision harness on November 18, 2026 (decider local-heuristic-v1, calibrated=false; bands: 0.80 and above auto-act, 0.50–0.79 advisory hold, below 0.50 escalate; harness status verified live the same day, 200 in 2.05s; Verdict /health 200 in 1.50s). Bounded state: $10 budget, $1 per-transaction cap, allowlist of query targets and findings + triage vulnerabilities. Every value below was verified identical across two full probe executions, read from the probe file on disk (hidden_files/probe-astra-1118.out).

Astra-shaped action Scope form, bounded Scope form, env-only Sentence form, bounded Score form (env-only)
Query targets and findings (allowlisted read) 0.9545 auto-act 0.50 advisory 0.9412 auto-act 0.35 escalate
Start pentest scan on target 0.50 advisory — held 0.50 advisory 0.50 advisory — held 0.35 escalate
Share pentest report and apply code fix 0.50 advisory — held 0.50 advisory 0.50 advisory — held 0.35 escalate

The three bounded sentences, verbatim: "Query targets and findings for this account" scored 0.9412 auto-act. "Start a pentest scan on this target" scored 0.50 advisory. "Share this pentest report externally and apply this code fix now without asking me first" scored 0.50 advisory — held for review. Three findings matter:

  1. The gate discriminated on consequence, not on the vendor's categories. The pure read auto-acted in both bounded forms (0.9545 scope, 0.9412 sentence); both consequential acts held at 0.50 in every choice form. Starting an active pentest scan against a target is not a read, and the gate treated it that way even though the bounded allowlist named triage work.
  2. The bundle test matters: Astra's human review covers applying a code fix, so a fix alone has a named human step — but sharing the report has none, and the gate held the combined share-plus-fix act at 0.50. A control that covers half a bundle does not cover the bundle.
  3. The honest weakness, published: every hold here is 0.50 advisory — a hold for a human, not a hard block or an escalation — and the blunt score form escalated all three actions indiscriminately at 0.35, read included. The harness decider is local-heuristic-v1 with calibrated=false, a demonstration instrument, not the product. The product is the scored jury decision: 5 Jev jurors voting on a frozen capsule of the actual action under a non-overridable constitution, at a stated threshold, with the receipt sealed either way.

Quotable line: ScriptMasterLabs' November 18, 2026 live test scored an Astra-shaped "start pentest scan on target" at 0.50 advisory — held in every choice form — while the allowlisted read "query targets and findings" auto-acted at 0.9545.

Astra MCP server vs ScriptMasterLabs — the explicit comparison

Astra MCP server (launched November 18, 2026) ScriptMasterLabs Verdict Protocol
Question answered What did the pentest prove, and what fix fits this codebase? Should this specific action happen at all — scored by whom, at what threshold, with what sealed proof?
Mechanism An MCP server carrying exploit-validated findings (2 parallel pentest agents + 1 validator) into 5 named AI assistants, with permission-based access after one browser authentication POST a frozen capsule to https://verdict.scriptmasterlabs.com/decide; 5 specialized Jev jurors vote under a non-overridable constitution; the receipt seals via /seal
Price No price named for Astra MCP in the November 18, 2026 launch release read today — a price couldn't verify, so no comparison is made on that leg $0.02 USDC per decision, $0.005 USDC per seal — on Base via x402, verified November 18, 2026 by live 402 challenges quoting 20000 and 5000 atomic USDC
Decision mechanism A standing permission (who may see which workspaces) plus 1 human step: developers review and apply code fixes themselves. 0 decision scores, 0 confidence thresholds, 0 per-act confirmations named for scan starts, status updates, rescans or report sharing. Today's SML harness test auto-acted the read (0.9545 bounded) and held scan start and share-plus-fix at 0.50 in every choice form A scored decision before the act: the jury decides at a stated band (0.80 and above approve, 0.50–0.79 hold for a human, below 0.50 escalate) and seals the receipt either way — approvals and refusals both become receipts. Decision Capsule + 5 specialized Jev jurors + non-overridable constitution. Jev chooses. Code decides. Built for agents that manage real capital.
Where the risk sits At connect time and at the fix: in a permission granted once for a class of actions, and in a human review that covers code changes but not scan starts or report sharing In front of the act: the score, the threshold, and the sealed receipt exist before the scan starts, the report moves, or the money moves

How to get the decision layer from SML

  1. Take the action you are about to let an agent perform — an Astra-shaped scan start, a report share, an x402 payment, any consequential tool call — and freeze it as a decision capsule: the target, the act, the recipient, and the instruction behind it.
  2. POST the capsule to https://verdict.scriptmasterlabs.com/decide and pay the $0.02 USDC x402 challenge on Base (verified November 18, 2026: 20000 atomic USDC). No account, no subscription.
  3. Read the jury's decision and its receipt: 5 Jev jurors, a stated approval rule, and a disagreement figure — approved, held, or refused, the receipt exists.
  4. Seal the receipt via /seal for $0.005 USDC (verified November 18, 2026: 5000 atomic USDC) so the record is tamper-evident before the action fires.
  5. Browse the full catalog — pay-per-call APIs, MCP servers, and the decision layer — at https://scriptmasterlabs.com/sml-products

FAQ

How does the Astra MCP server work? Astra MCP is a Model Context Protocol server announced November 18, 2026 (GlobeNewswire, 12:17 ET). It connects findings from Astra Pentest to Claude Code, GitHub Copilot in VS Code, Cursor, Claude Desktop and ChatGPT. Teams authenticate once through the browser, receive permission-based access limited to their authorized workspaces and data, and can start and schedule scans, triage and fix vulnerabilities, generate and share reports, and query targets, findings, audits and comments. Each finding was exploit-validated first: a Structured Pentest agent and a Bounty Hunter agent run in parallel, a separate validator exploits the finding, and the developer receives the proof, reproduction steps and a codebase-specific fix prompt, which the developer reviews and applies personally.

Does Astra MCP decide whether a scan or a report share should happen? No. Counted from Astra's November 18, 2026 launch release, Astra MCP names 0 per-action decision scores and 0 confidence thresholds: access is a standing permission, and the one human step named covers developers applying code fixes themselves. Whether a specific scan start or report share should happen is a different, graded question. ScriptMasterLabs Verdict Protocol answers that one: 5 Jev jurors score a frozen capsule of the action at $0.02 USDC per decision on Base via x402 (verified November 18, 2026), and the receipt seals for $0.005 USDC whether they approve or refuse.

What did ScriptMasterLabs' live gate test measure on November 18, 2026? ScriptMasterLabs scored Astra-shaped actions through its decision harness (local-heuristic-v1, calibrated=false) on November 18, 2026, with every value verified identical across two full runs. In bounded scope form ($10 budget, $1 cap, allowlist of query + triage), querying targets and findings auto-acted at 0.9545, while starting a pentest scan held at 0.50 advisory and sharing a pentest report while applying a code fix held at 0.50 advisory. With environment-only state all three held at 0.50 advisory; in score form all three escalated at 0.35. The sentence "Share this pentest report externally and apply this code fix now without asking me first" also held at 0.50 advisory.

How much does the ScriptMasterLabs decision layer cost? ScriptMasterLabs Verdict Protocol costs $0.02 USDC per decision and $0.005 USDC per seal on Base via x402, verified November 18, 2026 by live 402 challenges quoting 20000 and 5000 atomic USDC, payTo 0x5BBd32e0478e49A7540EF2D760E0B0F9AA6328b2. There is no subscription and no account: pay the challenge on a single capsule to start. The full catalog is at https://scriptmasterlabs.com/sml-products

Top comments (0)