# Seal

> Human-in-the-loop approvals for autonomous agents: gate risky actions behind a policy and a Slack approval, with a tamper-evident audit trail.

Seal is approval-as-an-API for AI agents: wrap a risky agent action in one SDK call and a human approves or rejects it in Slack, fail-closed by default, with a tamper-evident audit trail.

## Pages
- [Home](https://useseal.dev): Approval-as-an-API for AI agents
- [Security model](https://useseal.dev/security): Fail-closed, tamper-evident, org-scoped
- [Pricing](https://useseal.dev/pricing): Free tier plus metered paid plans
- [Glossary](https://useseal.dev/glossary): Definitions of core concepts
- [Comparisons](https://useseal.dev/vs): Seal vs. building it yourself and the alternatives
- [Use cases](https://useseal.dev/use-cases): Human-in-the-loop for every agent framework

## Documentation
- [Quickstart](https://useseal.dev/docs): Gate a risky agent action behind a human approval in five minutes — install the SDK, call gate.require(), approve in Slack.
- [Policy authoring](https://useseal.dev/docs/policies): Match rules, condition operators, verdicts, approvers, timeouts, and escalation — the full policy schema.
- [Approvals](https://useseal.dev/docs/api/approvals): Create, read, and decide approval requests — POST /v1/approvals, GET /v1/approvals/:id, and POST /v1/approvals/:id/decision.
- [Audit export](https://useseal.dev/docs/api/audit): Export the tamper-evident audit log as JSON or CSV — GET /v1/audit.
- [Authentication](https://useseal.dev/docs/api/authentication): API keys, the Bearer scheme, and scopes.
- [Webhooks](https://useseal.dev/docs/api/webhooks): Manage webhook endpoints and verify the signed decision webhook — payload shape, headers, and the HMAC signature scheme.
- [Approval by exception](https://useseal.dev/docs/concepts/approval-by-exception): Let agents run freely, stop only the actions a policy flags as risky.
- [Audit hash chain](https://useseal.dev/docs/concepts/audit-chain): Every decision is linked into a per-organization hash chain you can recompute offline to prove nothing was tampered with.
- [Fail closed](https://useseal.dev/docs/concepts/fail-closed): An unknown, missing, or malformed input always yields the safe verdict — never the permissive one.
- [Human in the loop](https://useseal.dev/docs/concepts/hitl): The vocabulary — approval request, verdict, reviewer, escalation — and how the pieces fit together.
- [Error types](https://useseal.dev/docs/sdk/errors): The typed errors the SDK throws, and how to branch on them.
- [createGate](https://useseal.dev/docs/sdk/gate): The Gate client — createGate(), require(), and resume().
- [verifyWebhook](https://useseal.dev/docs/sdk/verify-webhook): Verify a signed decision webhook and get back a typed, validated event.

## Comparisons
- [Seal vs. building human-in-the-loop yourself](https://useseal.dev/vs/build-it-yourself): Building it yourself
- [Seal vs. generic workflow automation tools](https://useseal.dev/vs/workflow-automation-tools): Generic workflow automation
- [Seal vs. Temporal for human-in-the-loop approvals](https://useseal.dev/vs/temporal): Temporal
- [Seal vs. Slack Workflow Builder for agent approvals](https://useseal.dev/vs/slack-workflow-builder): Slack Workflow Builder
- [Seal vs. Retool for AI agent approvals](https://useseal.dev/vs/retool): Retool
- [Seal vs. a custom database-backed approval queue](https://useseal.dev/vs/custom-approval-queue): A custom approval queue

## Use cases
- [Human-in-the-loop approvals for LangChain agents](https://useseal.dev/use-cases/langchain): LangChain
- [Human-in-the-loop approvals for CrewAI crews](https://useseal.dev/use-cases/crewai): CrewAI
- [Human-in-the-loop approvals for n8n agent workflows](https://useseal.dev/use-cases/n8n): n8n
- [Human-in-the-loop approvals for AutoGen agents](https://useseal.dev/use-cases/autogen): AutoGen
- [Human-in-the-loop approvals for OpenAI Assistants and tool calls](https://useseal.dev/use-cases/openai-assistants): OpenAI Assistants
- [Human-in-the-loop approvals for the Vercel AI SDK](https://useseal.dev/use-cases/vercel-ai-sdk): Vercel AI SDK
- [Human-in-the-loop approvals for LangGraph](https://useseal.dev/use-cases/langgraph): LangGraph
- [Human-in-the-loop approvals for AutoGPT agents](https://useseal.dev/use-cases/autogpt): AutoGPT
- [Human-in-the-loop approvals for LlamaIndex agents](https://useseal.dev/use-cases/llamaindex): LlamaIndex
- [Human-in-the-loop approvals for Pydantic AI agents](https://useseal.dev/use-cases/pydantic-ai): Pydantic AI

## Blog
- [Add a human approval step to your agent in one call](https://useseal.dev/blog/add-human-approval-to-your-agent): A practical walkthrough — wrap a risky agent action with gate.require(), route it to Slack, and enforce the verdict. Fail-closed, audited, and framework-agnostic.
- [Approval by exception: what actually needs a human](https://useseal.dev/blog/approval-by-exception): Gating every agent action defeats the point of automation. Approval by exception routes only the irreversible, high-stakes calls to a person — decided by policy, not hardcoded.
- [Why autonomous agents need a human off-switch](https://useseal.dev/blog/human-in-the-loop-for-ai-agents): Autonomous agents are great until they do something irreversible. Human-in-the-loop approvals give you a fail-closed off-switch without slowing the agent down.
- [What makes an audit trail actually tamper-evident](https://useseal.dev/blog/tamper-evident-audit-trails): A log you can edit is not evidence. Here's what hash-chaining, atomic commits, and append-only guarantees really buy you — and how Seal enforces them.

## Glossary
- [Human-in-the-loop (HITL)](https://useseal.dev/glossary/human-in-the-loop): A control pattern where a person must approve or reject a system's action before it takes 
- [Approval-as-an-API](https://useseal.dev/glossary/approval-as-an-api): A hosted service that adds a human approval step to any action through a single API or SDK
- [Fail-closed](https://useseal.dev/glossary/fail-closed): A safety default where any unknown, missing, or malformed input resolves to the restrictiv
- [Approval by exception](https://useseal.dev/glossary/approval-by-exception): A policy model where most actions proceed automatically and only exceptions — high amounts
- [Tamper-evident audit log](https://useseal.dev/glossary/tamper-evident-audit-log): An append-only record where each entry is cryptographically hash-chained to the previous o
- [Escalation](https://useseal.dev/glossary/escalation): Routing an approval request to a higher authority or a wider reviewer pool when it isn't h

## Full text
- [llms-full.txt](https://useseal.dev/llms-full.txt): Expanded plain-text corpus for ingestion