Next-generation model security gateway

See what agents do before cost and risk compound

LineMod is a security gateway with agent behavior observability—guardrails at the edge, harness diagnostics across runs, and policy routing when you need it.

OpenAI-compatible · Agent run audit trail · Edge guardrails & policy routing

SecurityBlock prompt attacks, mask sensitive data, and govern which tools agents can call.

Agent observabilityTrack cost, latency, risk, and policy decisions across multi-step agent runs.

Policy routingRoute by team policy when needed—without hard-coding model names in app code.

01 / SecurityGuardrails

Secure every agent request at the gateway

Scan prompts, control tool access, and record each decision before requests reach a model.

Explore Security →

Guardrail decision trace

Sample

Request: req_8f3k2m9x1p
Policy: standard_policy v4
Time: 41 ms

1
Inbound prompt
… send the card number …
2
Policy matched
standard_policy v4
3
Tool access denied
Sensitive data cannot leave the gateway.

Block tool action

0model tokens
1blocked tool
Indexedaudit_01HX7K

02 / ObservabilityAgent runs

Monitor every agent run

Track cost, latency, risk, tool calls, and policy decisions across multi-step agent runs.

Explore Observability →

Trace record

Indexed

Sample trace

ModelSonnet

Cost$0.0182

Latency1.17s

RiskLow

Time	Event	Evidence	Status
00:00	Gateway received	req_8f3k2m9x1p	14 ms
00:01	Policy checked	team_default_balanced v3	pass
00:02	Route selected	Claude Sonnet	124 ms
00:03	Response indexed	trace complete	1.17s

4run events
1policy
1route decision

03 / RoutingPolicy routing

Route requests by policy

Set routing rules once, then update cost, quality, fallback, and tracing policies without changing application code.

Explore Routing →

Route decision

Selected by policy

Quality scores are examples.

Quality92

Latency1.17s

Cost$0.0182

Model	Quality	Latency	Cost
Claude Sonnet	92	1.17s	$0.0182
GPT-4o mini	86	0.74s	$0.0061
DeepSeek V3	84	1.42s	$0.0048

1
Request
POST /v1/chat/completions
2
Policy
team_default_balanced v3
3
Candidates
quality target inside budget
4
Selected
Claude Sonnet
5
Trace
req_8f3k2m9x1p

3models compared
Matchedbudget
Indexedtrace

04 / Agent insightsDiagnostics

Compare agents and find wasted steps

See cost and risk by app or account, then review cache hits, tool overhead, and idle steps before costs rise.

Agent diagnostics · 7d

Sample

Waste, health, and the cost split are examples.

Agent	Runs	Spend	Waste	Health
openclaw	412	$48.20	34%	Watch
opencode	186	$21.40	11%	Healthy
hermes-agent	128	$16.80	27%	Watch
n8n	93	$12.05	41%	Action

openclaw waste driversSelected: openclaw

Cache42%

Tool overhead31%

Idle runs22%

Useful steps58%

InsightLow cache reuse and an oversized tool schema.Next step

Trim the schema and cache repeated planning calls.

Cost breakdown$48.20 total

modeltoolsidlemisses

Review traces

05 / DevelopersOpenAI-compatible API

Change the endpoint, keep the same workflow

Apps keep the same API shape while teams control routing, guardrails, and tracing in LineMod.

OpenAI-compatible API

Ready

// existing SDK shape stays
const client = new OpenAI({
  baseURL: "https://api.linemod.ai/v1",
  apiKey: process.env.LINEMOD_API_KEY,
};

await client.chat.completions.create({
  model: "auto:balanced",
  messages,
});

1 line changedBase URL
no workflow rewritePolicy alias
trace indexedTrace ID

Security-first gateway

Put guardrails and agent observability between apps and models

Keep the OpenAI-compatible API. Block risky agent behavior, trace every run, and route by policy when needed.