faultgate 0.3.1 · fixes from the first outside test, 180 real tau2-bench runsRead the changelog →
SERVICES

Services

Use the tools yourself for free, or have us run the work with you.

01 / ENGAGEMENT OPTIONS

Hands-on engineering engagements led by the founder.

Readiness Review

$7,5005 business days
WHO IT IS FOR

Teams deploying multi-tool or autonomous agents to production who need an objective release verdict and evidence pack for security or leadership reviews.

WHAT YOU GET
  • We write your policy.json rules for domain-specific constraints and business invariants
  • Import your traces from your current tool (OTLP, Langfuse, LangSmith, Braintrust, Phoenix)
  • 3+ baseline runs to measure your empirical noise band and operational variance
  • Verdict + audit pack with the SHA-256 of every input recorded, failing traces, and local reproduction scripts
Talk to the founder →

Custom eval suite build

Contact usScope and price agreed per engagement
WHO IT IS FOR

Teams building agents whose real-world tasks require custom scenarios, golden answers, and deterministic checks instead of an LLM judge.

WHAT YOU GET
  • Scenarios built from your agent's real tasks and workflow requirements
  • Deterministic checks (golden answers, business rules) instead of an LLM judge wherever the failure is structural
  • Baseline over repeated runs (empirical noise band, pass^k joint reliability)
  • Wired directly into your CI pipeline with faultgate exit codes
  • The suite lives in your repo with complete local reproduction scripts
Talk to the founder →

Tool-poisoning red-team

Contact usScope and price agreed per engagement
WHO IT IS FOR

Teams connecting agents to external or third-party MCP servers and tools who need to verify runtime protection against prompt injection and tool manipulation.

WHAT YOU GET
  • Attack battery modelled on the three paradigms measured in FAULTLINE Publication 02 (tool-description injection, function hijacking, parameter injection) plus document-text injection categories (P9)
  • Paired runs with and without defences to measure exact intervention efficacy
  • Report with attack success rates and 95% Wilson confidence intervals
  • Recommended runtime policy (mcp-guard rules / faultgate policy.json)
Talk to the founder →

Ongoing reliability retainer

Contact usScope and price agreed per engagement
WHO IT IS FOR

Teams operating production agents who need continuous verification across upstream model swaps, prompt iterations, and tool schema updates.

WHAT YOU GET
  • Periodic re-baselining when underlying foundation models or prompts change
  • Review of release verdicts and investigation of borderline or inconclusive runs
  • Model-retirement migration checks and benchmark comparisons
  • Rule-pack maintenance as your tool schemas and APIs evolve
  • Priority support channel with direct founder access
Talk to the founder →

02 / PROCESS

Transparent 4-step delivery lifecycle.

How an engagement works.

We work directly with your engineering team to establish reproducible evaluation gates.

01

Talk

30 min with the founder to review your agent architecture, current tracing setup, and evaluation goals.

02

Scope

Written scope, timeline, and fixed price agreed before any work begins. No open-ended hourly billing.

03

Deliver

Executed in your environment where possible; strict data handling parameters agreed before we start.

04

Handover

Everything stays in your repo: test suites, golden datasets, policy rules, and reproduction scripts.

03 / FAQ

Common questions about services and open-source licensing.

Do I need to buy anything to use the tools?

No. faultgate (pip install faultgate) and mcp-guard (pip install mcp-integrity) are MIT and free; the FAULTLINE Evals workbench is an MIT repo you clone and run (see /platform/evals).

Do you need our data?

Data handling parameters are agreed per engagement before work begins. Tools run locally in your environment where possible, and traces never need to leave your network.

Can services be combined?

You can combine them. For example, a Readiness Review first, then a retainer to keep the baseline current.

What does 'Contact us' cost?

Every engagement has a fixed, agreed written scope and price before any work begins. We discuss your architecture on a 30-minute founder call and provide a transparent written proposal.

WORK WITH THE FOUNDER

Talk to the founder to discuss your agent's reliability.

Schedule a 30-minute working call to review your failure modes, policy rules, and evaluation setup.

Talk to the founderSee pricing →