Legal

One approved prompt.
Not fifty personal copies.

Versioned review templates standardize the firm's prompts, matter-level budgets keep client billing clean, and an append-only audit trail records every change for the file.

Talk to Sales See the controls

Pass-through gateway · prompts never train nRouter · configurable data policy

All retainedPrompt versionsEvery template revision kept
3Budget scopesFirm · matter/team · per-key
Append-onlyAudit trailEvery action recorded
ConfigurableData policyDown to zero-content logging

Review workflow

Standardize the review prompt. Keep the history.

A practice group drafts and A/B tests a contract-review prompt, saves the approved version as a server-side template, and every associate runs that exact version. Each revision is retained, every change lands in the append-only audit trail, and the cost of each review bills to the matter team that ran it.

Server-side templates: no prompt copy lives on a laptop
Every version retained for rollback and comparison
A/B test two variants on real matters before adopting one firm-wide
Matter team budgets make AI cost a clean disbursement line: 402 on a tripped hard cap

How prompt templates work

Step 01 — draft + version

Approved template saved

Practice group tests variants · every revision retained

Step 02 — run

Associates run the firm version

Cost lands on the matter team budget, per key

Step 03 — record

Audit trail, append-only

Every template change captured with actor and diff

Legal Enterprise Architecture

Privilege-Preserving Request Pipeline: From Firm DMS to Model Egress

Every legal analysis and e-discovery prompt traverses in-memory privilege filters, matter-isolated accounting, and enterprise zero-retention routes before model execution.

Enterprise request lifecycle: client → gateway → guardrail filter → smart router → model provider

Legal Tech / DMS App
POST /v1/chat/completions
Contract drafting, discovery review, matter research.
Gateway Preflight & Auth
:4000 · Matter Scope
Virtual key auth, matter budget check, ethical walls.
Guardrail Filter
Privilege & PII Masking
Work-product privilege, PII, and client secrets shielded.
Smart Router & Optimizer
Matter Routing · Latency
50–70% cost savings via discovery vs reasoning tiering.
Model Providers
Zero-Training Endpoints
99.99% uptime during filings; client data never trains models.

50–70%

Cost Savings

Via matter tier routing

99.99%

Filing Uptime

Multi-provider fallback

< 1 ms

Gateway Overhead

In-memory routing proxy

ABA 1.6

Confidentiality

Privilege safeguards

Stage 01 — Matter Ingress & Ethical Walls

Matter-Scoped Virtual Keys & Isolation

Requests authenticate with matter-specific virtual keys (sk-nrouter-matter-...). The gateway enforces strict ethical walls and billing boundaries, validating that matter billing codes match open client files before allowing execution.

Stage 02 — Inline Privilege Protection

Attorney-Client Privilege & Confidentiality Filters

Payloads pass through inline guardrail scanners configured for ABA Model Rule 1.6 compliance. Client identity, case numbers, and settlement terms are masked or encrypted before egress, ensuring privileged communications remain fully confidential.

Stage 03 — Workload Tier Routing (50–70% ROI)

Discovery Batching vs. Complex Brief Synthesis

Bulk document classification during e-discovery routes to high-throughput, low-cost flash models, saving 50–70% compared to monolithic frontier model usage. High-stakes appellate analysis and contract clause negotiation dynamically route to top-tier reasoning models.

Stage 04 — Enterprise Zero-Training Egress

Court Filing Deadline Failover (99.99% SLA)

nRouter routes exclusively to enterprise zero-retention provider contracts where client inputs are never used to train public or foundational models. Multi-provider fallback guarantees continuous availability during midnight court filing deadlines.

Stage 05 — Client Disbursement Accounting

Automated Matter Ledger & Audit Records

Inference concludes with exact list-price micro-cost settlement attributed to the matter file code. Costs export directly to law firm practice management systems (Clio, Elite, Aderant) as clean, defensible client disbursement line items.

Firm Integration

Integrating Matter-Specific Keys and Privilege Filters

Deploy firm-wide review workflows using standard OpenAI SDK clients. Route prompts with client matter tags and privilege filters using HTTP headers.

Installpip install openai

1	`# Cache: enabled (org default). Pass nrouter_cache: false to skip.`
2	`from openai import OpenAI`
3	`import os`
4
5	`client = OpenAI(`
6	`api_key=os.environ["NROUTER_API_KEY"],`
7	`base_url="https://api.nrouter.ai/v1",`
8	`)`
9
10	`response = client.chat.completions.create(`
11	`model="gpt-5.4-mini",`
12	`temperature=1,`
13	`max_completion_tokens=1024,`
14	`messages=[`
15	`{"role": "user", "content": "Hello! What models do you support?"},`
16	`],`
17	`extra_body={`
18	`# "nrouter_cache": False, # Uncomment to skip cache`
19	`},`
20	`)`
21
22	`print(response.choices[0].message.content)`

Header Configuration & Policy Spec

# Legal Matter Integration (OpenAI Python SDK)
from openai import OpenAI

client = OpenAI(
    base_url="https://api.nrouter.ai/v1",
    api_key="sk-nrouter-matter-2026-lit-0842",
    default_headers={
        "x-nr-matter-id": "2026-LIT-0842-ACME",
        "x-nr-guardrails": "privilege-protect,pii-mask",
        "x-nr-data-policy": "zero-content",
        "x-nr-budget-scope": "matter-lit-0842"
    }
)

response = client.chat.completions.create(
    model="nrouter/auto",  # Automatically selects tier
    messages=[
        {"role": "system", "content": "You are a legal review assistant. Analyze indemnity clauses for cross-liability."},
        {"role": "user", "content": "Review Section 14.2 of the attached Master Services Agreement."}
    ],
    extra_body={
        "fallback_models": ["anthropic/claude-3-5-sonnet", "openai/gpt-4o"],
        "max_cost_limit": "0.10"
    }
)

Matter-Level Budget Caps (HTTP 402)

Cap AI disbursements per client matter. If an associate or automated e-discovery job reaches the client’s authorized budget cap, execution halts with clean 402 errors.

Zero Provider Training Guarantee

All provider traffic traverses enterprise zero-retention contracts. Client legal drafts, merger agreements, and patent disclosures never train foundation models.

Versioned Prompt Templates for Briefs

Centralize firm-approved contract review checklists, deposition outlines, and memo formats. Version controls track modifications with full attorney attribution.

Also on every plan

Core enterprise capabilities on every plan

Confidentiality controls on every request

PII redaction runs inline, and a zero-content or metadata-only data policy keeps client content unlogged.

Matter-level budgets

Hard and soft caps per firm, matter team, and key. Per-key spend breakdown for client cost recovery.

Encryption + tenant isolation

AES-256 at rest, TLS 1.2+ in transit, and Postgres Row-Level Security isolating every firm and team.

Confidentiality — honest status

What we can state plainly about your data.

nRouter is a pass-through gateway. Prompts are forwarded to the provider you route to; they are not used to train nRouter.
Content stays out of our logs when you set a zero-content or metadata-only data policy, and PII redaction strips identifiers before a request is forwarded.
SOC 2 Type II is in progress (status as of August 2026); a DPA is published and signable. We will confirm the provider data-handling terms for the models in your routing config.

Need the DPA and a data-handling walkthrough for a conflicts or risk review? security@nrouter.ai will take your team through it.

Legal questions, answered

How does nRouter protect attorney-client privilege and ABA Model Rule 1.6 confidentiality?

nRouter operates as an enterprise pass-through gateway with zero-content data policies and in-process PII/privilege redaction. Your client data is evaluated in memory, forwarded across encrypted mTLS channels to zero-retention provider endpoints, and never stored in plain text.

Upstream providers are legally bound by enterprise agreements barring customer data from training foundation models.

How do matter-level budgets prevent client billing disputes?

Every virtual key or request header can be mapped directly to a client matter number (e.g. 2026-LIT-0842).

Spend is tracked at micro-cent precision from provider-reported token usage. If a matter reaches its authorized budget ceiling, further calls return HTTP 402.

Billing reports export cleanly into practice management tools (Aderant, Elite 3E, Clio) with verifiable timestamps and token breakdowns.

Yes. Prompt templates live in a central, versioned repository.

Instead of associates copying prompts into private chat tools, applications call templates by ID with dynamic variables. Every template update is logged with author, timestamp, and diff, allowing managing partners to maintain rigorous quality control across all offices.

Can client files be restricted to specific sovereign jurisdictions?

Yes. Data residency can be pinned to US sovereign cloud regions, EU zones (for cross-border GDPR compliance), the UK, Canada, Australia, or Singapore.

Prompts, cached contexts, and routing metadata never transit outside the designated jurisdiction.

How does multi-provider failover prevent missed court filing deadlines?

Court filing deadlines cannot tolerate provider outages or API downtime. If your primary model endpoint degrades or encounters rate limits, nRouter transparently retries through alternate providers (Azure OpenAI, AWS Bedrock, Anthropic) within 50 ms, ensuring discovery jobs and draft briefs complete on schedule.

Explore industry solutions

Tailored AI gateway controls for every vertical

Finance Healthcare Retail & e-commerce Public sector Education All solutions →