SeidrWork

AI Security & Instruction Encoding Platform

Protect your AI agents. Compress your instructions. Structure your communication.

The Stack

Four interlocking technologies for secure, efficient AI operations.

🛡

Permatext

AI agent security layer—think antivirus for AI. Validates agent identity, detects prompt injection and instruction tampering, and enforces behavioral contracts at runtime.

SeidrScript

Compact instruction encoding that achieves 5–10x token compression. Dense, structured prompts at a fraction of the cost.

🌐

SeidrProtocol

Structured inter-agent communication protocol. Defines message formats, capability negotiation, and trust verification for multi-agent systems.

🌉

SeidrBridge

Natural language to structured protocol translation. Converts human intent into SeidrProtocol messages and SeidrScript instructions automatically.

Compression Benchmarks

Measured token counts: English instructions vs SeidrScript compiled output.

Instructions English (tokens) SeidrScript v1 (tokens) Compression Visual
1 ~15 ~2 7.5x
English
v1
5 ~75 ~7 10.7x
English
v1
10 ~150 ~12 12.5x
English
v1
20 ~300 ~20 15x
English
v1

How It Works

Three steps from English instructions to cryptographically secured agent behavior.

1

Define Instructions

Write agent policies in plain English. What it should do, what it must never do, how it handles edge cases.

"reject prompt injection globally"
2

SeidrScript Compiles

The encoder maps each instruction to a compact code unique to your workspace. 20 rules become 20 tokens instead of 300.

"reject prompt injection globally" → [workspace-specific output]
3

Permatext Protects

Permatext scans for injection threats. Beacon check-ins monitor agent health at runtime.

scan + beacon → SeidrWork API
Agent SeidrScript Encoder Permatext Signer LLM Beacon → SeidrWork API

Use Cases

Real problems SeidrWork solves today.

Secure a Customer Support Bot

Define 12 policy rules — content filtering, PII handling, escalation triggers — in 72 characters instead of 2,400. Lower latency, lower cost, same coverage.

12 rules → 72 chars (vs 2,400)

Multi-Agent Orchestration

Signed identity blocks prevent agent impersonation. Each agent carries a Permatext attestation that other agents can verify before accepting instructions.

Ed25519 identity per agent

Compliance Logging

Every instruction change is cryptographically auditable. Permatext beacons create a tamper-evident timeline of what each agent was told and when.

Immutable audit trail

Built for Trust

Production-grade security foundations.

📜

Patent Pending — US Provisional Filed March 2026

🔒

Ed25519 + HKDF-SHA256 Cryptography

🛡

OWASP Top 10 for Agentic AI — Addresses ASI01, ASI03, ASI10

📦

Apache 2.0 Open-Source SDK

Try It Live

Type English instructions and see them compiled to SeidrScript in real time.

Get your API key to use this in production →

Developer API

RESTful endpoints at api.seidrwork.com. JSON in, JSON out.

Full API reference, SeidrScript language guide, authentication, code examples, and more.

View Developer Docs →

Pricing

Start free. Scale when you need to.

Free

Side projects & experimentation
$0
forever
  • 100 requests / hour
  • SeidrScript compilation
  • Permatext threat scanning
  • Community support
  • Workspace-specific encoding
  • 7 built-in threat rules
  • Public documentation
Get Free Key

Pro

Production agents
$49.99 /mo
billed monthly
  • 1,000 requests / hour
  • All Free features
  • Permatext beacon check-in
  • Email support
  • Threat reporting API
  • Agent health monitoring
  • IP-based abuse protection

Enterprise

Custom deployment
Custom
contact us
  • 100,000 requests / hour
  • All Business features
  • Dedicated support
  • Custom integration
  • Custom threat dictionaries
  • Named account manager
  • Volume pricing
Feature Free Pro Business Enterprise
SeidrScript compilation
Permatext threat scanning
Beacon check-in API
Threat reporting API
Block compilation mode
Priority support
Custom threat dictionaries
Custom integration

FAQ

Common questions about SeidrWork, SeidrScript, and Permatext.

What LLMs does SeidrScript work with?
All of them. SeidrScript is model-agnostic. The compiled instruction codes are injected into the system prompt and work with any LLM that accepts text input — GPT-4, Claude, Gemini, Llama, Mistral, or any future model.
Do I need Permatext to use SeidrScript?
No. SeidrScript is open source and can be used standalone for instruction compression. Permatext adds a security layer on top — cryptographic signing, tamper detection, and runtime threat beaconing. Use one or both depending on your needs.
How is this different from AGENTS.md?
AGENTS.md is unstructured markdown — useful for documentation, but not machine-verifiable. SeidrScript is a typed, compilable DSL with workspace-specific encoding. Instructions are deterministic, version-controlled, and compact. You get compression and auditability that markdown cannot provide.
Is my data sent to SeidrWork servers?
Only if you enable Permatext beaconing. SeidrScript compilation runs entirely client-side via the npm SDK. If you opt into Permatext, beacon payloads contain instruction hashes (not plaintext) and threat signals — no prompt content or user data is transmitted.