AI agents shouldn't make unverified commitments on your behalf. TermLens places a deterministic verification gate between WebMCP tools and your contract state, ensuring every extraction is grounded in real source text.
This Master Services Agreement is entered into as of May 1, 2026 between the parties named below.
Each party agrees to maintain the confidentiality of the other's proprietary information.
This Agreement shall automatically renew for twelve (12) months.
Either party may terminate this Agreement immediately upon written notice in the event of a material breach.
Client shall pay the sum of $48,000 per year in quarterly installments.
Notice of non-renewal must be given at least thirty (30) days before the term ends.
This Agreement is governed by and construed in accordance with the laws of the State of Delaware.
Neither party shall be liable for indirect or consequential damages.
"Automatically renew ... unless notice of non-renewal at least thirty (30) days before the term ends."
Standard LLM wrappers accept model extractions at face value. When an agent hallucinates a payment date or misinterprets a liability limit, the error quietly propagates into your record.
TermLens fundamentally alters this paradigm by separating proposal from execution. Agents can analyze documents and stage recommendations, but they are architecturally blocked from saving state directly.
Every proposal is subjected to local string matching and confidence scoring before reaching the staging area. If a proposed value lacks a direct quote in the contract text, it is flagged immediately for human review.
Three steps. Your agent proposes the breakdown. You hold complete control over execution.
Drop contract text directly into browser memory. Your documents are parsed locally and never uploaded to an external server.
Proposed terms stage in a split comparison view, pairing extracted values alongside verbatim source sentences and plain-language interpretations.
Approve or reject individual terms with 1-click. Toggle write-locks to revoke agent permissions, and export verifiable Markdown (.md) reports.
WebMCP tool calls can propose modifications, but write-locks enforce strict human authorization before any term is persisted to memory.
Every term is programmatically evaluated against source sentences in the text. Ungrounded values trigger visual confidence warnings automatically.
Maintain full accountability. Track all agent proposals and human approvals in exportable JSON audit logs and Markdown summary reports.
Open the app. Let your agent do the heavy lifting. Keep complete oversight over every term.
Open the app