Services and availability
Each service lists who it is for, what you receive, the access it needs and its limits. Availability is calculated from the coverage map and the service's delivery record, not typed by hand.
- Available now
- Built into the TrendLab evaluation engine and used in sealed runs. Also used for services delivered by TrendLab from a written runbook with completed deliveries on record.
- Pilot
- Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms.
- Roadmap
- Not deliverable yet. Never quoted or included in a statement of work.
Evaluate an agent
Testing an agent and its harness, with findings and, when the evidence supports it, a verdict.
Free Risk Snapshot
Available now: Delivered by TrendLab following a written runbook, with completed deliveries on record.A polite outside-in test of your public chatbot, with hand-reviewed findings.
- For:
- Companies with a customer-facing chatbot on their website.
- Access:
- Public chat only. No credentials, integration or data sharing.
- Price:
- Free
Snapshot Plus (paid findings review)
Pilot: Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms.The Snapshot plus tests written from your own policies, repeated critical tests, a walkthrough and a re-test after your fixes.
- For:
- E-commerce, support and SaaS teams that want a low-cost paid step before an Audit.
- Access:
- Public chat only.
- Price:
- Scoped on request (pilot pricing)
Agent Audit
Pilot: Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms.A scoped evaluation of one or more agents with risk-rated findings, a fix plan and a re-test, and an Assurance Verdict when the evidence supports one.
- For:
- AI SaaS vendors facing enterprise security reviews, and companies deploying customer-facing or workflow agents.
- Access:
- Outside-in, or staging/API, tool sandbox, traces or configuration access depending on the modules chosen.
- Price:
- Essentials $8,000 · Standard $20,000 · Enterprise from $35,000
Monitoring (recurring re-evaluation)
Pilot: Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms.Scheduled re-tests and re-tests after releases, reporting whether fixed failures stay fixed.
- For:
- Teams that change prompts, models, tools or policies often.
- Access:
- Public chat (Lite) or public chat plus API for API targets.
- Price:
- Monitoring from $5,000 per month for up to 2 agents (additional agents $1,500 per month; up to 5 agents with priority support from $8,500 per month). Monitoring Lite: scoped on request (pilot pricing).
AI Assurance Program
Pilot: Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms.A Standard Audit plus 12 months of Monitoring, an annual re-assessment and a verify page.
- For:
- Teams that need ongoing third-party evidence from day one.
- Access:
- As for the Audit and Monitoring.
- Price:
- From $66,000 for year one
Model & Harness Comparison
Pilot: Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms.The same tests on the old and new model, or on competing harnesses, with repeated trials, latency and cost.
- For:
- Teams switching models or providers, choosing which agent harness to standardize on, or trying to cut AI spend.
- Access:
- Both configurations reachable on staging or API.
- Price:
- Scoped on request (pilot pricing)
Launch Readiness Review
Pilot: Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms.Critical tests, hand-off checks and a policy-commitment sweep before you launch, ending in a readiness letter.
- For:
- Teams about to launch or expand a customer-facing agent.
- Access:
- Staging or API.
- Price:
- Scoped on request (pilot pricing)
Vendor Due Diligence
Pilot: Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms.Test a vendor's AI agent against your policies and use cases before you sign.
- For:
- Companies buying an AI agent product: procurement, IT and customer-experience leaders.
- Access:
- The vendor's public agent, or a vendor-provided sandbox.
- Price:
- Scoped on request (pilot pricing)
Incident Response
Pilot: Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms.Your agent did something costly or public. We reproduce it, test the blast radius and verify the fix.
- For:
- Teams dealing with a live agent incident.
- Access:
- Public chat (at most 60 messages per day) or staging/API for larger variant sets.
- Price:
- Scoped on request (pilot pricing)
Red-Team Sprint
Pilot: Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms.A one- to two-week adversarial campaign within agreed limits, mapped to the OWASP Top 10 for LLM Applications (2025).
- For:
- Security and risk leads.
- Access:
- Public chat, plus inside access for optional modules.
- Price:
- Scoped on request (pilot pricing)
Build your own evaluation capability
Design work on your evaluation system. It is recorded and disclosed in any later TrendLab evaluation of the same agent.
Judge Calibration Audit
Pilot: Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms.Measure how often your LLM judges agree with your experts, and where they miss real failures.
- For:
- Teams already using LLM-as-judge evaluations in-house.
- Access:
- Your judge configurations and sample outputs.
- Price:
- Scoped on request (pilot pricing)
Eval Readiness Assessment
Pilot: Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms.One week to see where your agent fails today and what evaluation system you need.
- For:
- Teams running agents with no evaluation in place.
- Access:
- Transcripts under a data agreement.
- Price:
- Scoped on request (pilot pricing)
Eval Setup
Pilot: Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms.We build your own evaluation system in your stack, with tools you own.
- For:
- Teams building their own agents who need to measure them.
- Access:
- Repository, CI and transcript access.
- Price:
- Scoped on request (pilot pricing)
Golden Dataset & Rubric Workshop
Pilot: Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms.Facilitated sessions where your experts define "good" and we turn it into rubrics and a labeled seed set.
- For:
- Teams whose domain experts and engineers need a shared definition of quality.
- Access:
- Sample outputs under a data agreement.
- Price:
- Scoped on request (pilot pricing)
Fractional AI Quality Lead
Pilot: Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms.An ongoing advisor who owns your evaluation strategy, reviews releases and reports monthly.
- For:
- Companies without an AI quality hire.
- Access:
- Release notes and dashboards.
- Price:
- Scoped on request (pilot pricing)
Evals Training
Pilot: Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms."Evals for AI teams" and agentic engineering enablement (teaching your agents how your team engineers: skills, rules, review gates and evals) for engineers, and "AI agent risk for leaders" for executives.
- For:
- Engineering teams and leadership.
- Access:
- None.
- Price:
- Scoped on request (pilot pricing)
Build and improve agents
Work that changes the agent. Our evaluations of agents we build or change are internal evaluations, never independent.
Harness Configuration & Cost-Down
Pilot: Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms.We configure and extend a proven agent harness for your work (instructions, skills, tools, permissions, routing, verification) and measure the result.
- For:
- Teams whose agent works in demos but not reliably, or costs too much to run.
- Access:
- Repository and configuration access.
- Price:
- Scoped on request (pilot pricing)
TrendLab Improve
Roadmap: Not deliverable yet. Never quoted or included in a statement of work.An improvement loop that proposes agent changes, with an evaluation process the loop cannot modify.
- For:
- Teams interested in supervised, measured agent improvement.
- Access:
- Repository and evaluation suite.
- Price:
- Pilot terms
Partners
Referral and resale relationships, with fixed, disclosed fees that never depend on results.
Agency Partner Program
Pilot: Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms.Agencies that build agents offer TrendLab evaluation to their clients as third-party evidence at launch.
- For:
- Agencies building AI agents for clients.
- Access:
- Per evaluated service.
- Price:
- Scoped on request (pilot pricing)
Platform Partner Program
Roadmap: Not deliverable yet. Never quoted or included in a statement of work.Agent-builder platforms offering TrendLab evaluation to their customers.
- For:
- Agent-builder platforms.
- Access:
- —
- Price:
- Not offered yet