Blog · writing from the team
Notes on the runtime, the taxonomy, and the category we are building. Updated as we ship.
2026-04-28
The Open Intent now covers 29 domains spanning digital enterprise operations and physical / embodied processes. Same code structure, same Apache 2.0 license, same stable family codes. Manipulation, locomotion, sensing, manufacturing, autonomous vehicles, surgical robotics, agriculture, construction, hazardous environments, energy, mining, logistics, aerial systems, and embodied AI safety are now first-class.
2026-03-22
How much does an unauthorized AI action cost? The math makes the case for runtime authority better than any feature comparison.
2026-03-21
A step-by-step tutorial for adding authorization and audit to any LangChain agent using the Intended Python SDK.
2026-03-20
AI agents are in production at the majority of Fortune 500 companies. The regulatory landscape is catching up fast. Here is what security leaders need to understand.
2026-03-20
AI agents are governed by permission systems designed for humans. That is a problem. Permission asks whether an identity can access a resource. Authority asks whether an action should happen. The difference matters when AI agents are making thousands of decisions per hour.
2026-03-19
We ran 5 red team agents against our own platform. Here is what they found, and how we fixed every issue.
2026-03-18
There is no common language for describing what AI agents do. The Open Intent changes that. Open Intent is an open taxonomy — now 29 domains and 173 categories — for classifying AI agent actions. Apache 2.0 licensed.
2026-03-18
Every company deploying AI agents faces the same question — how do you trust autonomous software to act on your behalf?
2026-03-17
When your authorization system goes down, do AI agents keep executing? The choice between fail-open and fail-closed is the most important architectural decision in AI agent governance. Intended is fail-closed at every boundary.
2026-03-15
The Model Context Protocol is becoming the standard for AI agent tool use. But MCP has no built-in authorization. Here is how to add policy-based governance to every MCP tool call with the published Intended MCP middleware (a few lines of Python). A zero-config gateway proxy is in private preview.
2026-03-14
Audit logs tell you what happened. Cryptographic proof tells you what happened and proves it mathematically. Intended produces Ed25519-signed authority tokens, HMAC evidence bundles, and a hash-chained ledger for every decision.
2026-03-12
Intended is an AI-native company. Our AI agents handle support triage, billing operations, platform health, and more -- all governed by our own Authority Engine. Here is how we built it and what we learned.
2026-03-10
OPA is a great policy engine for infrastructure. But AI agents need more than policy evaluation. They need intent understanding, risk scoring, domain intelligence, and cryptographic proof. Here is the migration path.
2026-03-10
Defining the authority runtime category and what makes it different from authorization, policy engines, and access control.
2026-03-09
Real failure scenarios when AI agents operate without governance -- unauthorized purchases, data leaks, infrastructure damage, and why governance matters.
2026-03-08
A complete walkthrough of the 14 digital Open Intent domains that classify the actions an AI agent takes inside an enterprise. From software development to executive operations, here is the taxonomy that makes AI governance possible. (Open Intent v2 expands this to 29 domains and 173 categories across digital and physical operations.)
2026-03-08
RBAC was designed for humans clicking buttons. AI agents need intent-aware authorization that understands context, velocity, and risk.
2026-03-07
Technical deep-dive on the Intended decision pipeline -- how the deterministic authority decision meets a p50-under-80ms / p99-under-250ms SLO, with the token-validation path faster still (p50 under 50ms).
2026-03-06
A $5000 payment in FinOps vs. a test payment in sandbox -- same action, completely different risk. How domain intelligence makes governance accurate.
2026-03-06
Open-sourcing Open Intent was a strategic decision. We studied Databricks, Confluent, and HashiCorp to understand when open-sourcing a foundational technology creates more value than keeping it proprietary. Here is what is open, what is commercial, and why.
2026-03-05
Category-defining launch post for deterministic AI execution authority.
2026-03-05
What auditors look for, what most systems provide, and what Intended provides -- hash chains, evidence bundles, and independently-verifiable decision receipts.
2026-03-04
Deep dive into Ed25519 signing, nonce policy, and adapter verification flow.
2026-03-04
Kubernetes RBAC controls who can do what in a cluster. AI agents need governance beyond RBAC — intent classification, risk scoring, and cryptographic proof. Here's the admission-controller approach, and where it stands today.
2026-03-04
Applying zero-trust principles to AI agent operations -- never trust, always verify, always prove.
2026-03-03
Step-by-step tutorial with full code showing how to protect PydanticAI agent tools with Intended authority checks.
2026-03-03
Governance dashboards report controls, while runtime authority enforces them.
2026-03-02
Build a connector that verifies tokens and emits audit metadata.
2026-03-02
Tutorial walkthrough for wrapping OpenAI Agents SDK tools with Intended authority checks for production-grade governance.
2026-03-01
Transparent factor-level scoring for every authority decision.
2026-03-01
A 10-point evaluation framework for CTOs comparing AI governance solutions -- what to look for, what to avoid, and what questions to ask.
2026-02-28
Case studies of fail-open disasters and why Intended chose fail-closed as the only safe default for AI agent authorization.
2026-02-27
How Intended handles escalation workflows -- single approver, multi-party approval, delegation chains, and time-bounded authorization.
2026-02-26
Comprehensive guide to MCP security -- what MCP lacks in authorization, why it matters, and how Intended fills the gap.
2026-02-25
Industry-specific governance for financial services -- payment approvals, trading operations, regulatory compliance, and Intended's FinTech domain pack.
2026-02-24
Industry-specific governance for healthcare -- patient data access, clinical decision support, HIPAA considerations, and the healthcare domain pack.
2026-02-23
Industry-specific governance for DevOps -- deployment gates, infrastructure changes, incident response automation, and the infrastructure domain pack.
2026-02-22
Developer guide for creating organization-specific governance models with Intended domain packs -- from intent mappings to risk models.
2026-02-21
The vision for managing Intended policies as code with Terraform — and where the provider actually stands today (on the roadmap, not yet shipped).
2026-02-19
Engineering hours, maintenance burden, compliance gaps, and why buying AI agent authorization beats building it in-house.
2026-02-18
Enterprise procurement teams evaluate AI governance platforms against a specific checklist. SOC 2, DPA, SLA, uptime guarantees, data residency, and support tiers are table stakes. Here is what you need to pass the procurement gauntlet.
2026-02-17
OPA and Cedar are excellent policy engines. But building an AI governance platform on top of them requires solving the other 80 percent yourself. Here is an honest comparison of what they give you and what is missing.
2026-02-16
When an AI agent says it wants to do something, that request is natural language. Before governance can happen, that language must become structured data. Here is how Intended's intent compiler works, from raw text to classified intent.
2026-02-15
Authority tokens are cryptographic proof that an AI agent was authorized to take an action. But what stops an agent from using the same token twice? Nonces, TTLs, and single-use enforcement. Here is how Intended prevents token replay attacks.
2026-02-14
Most AI security tools protect one perimeter. But AI agents operate across four distinct perimeters -- ingestion, evaluation, execution, and audit. If you only secure one, you have three gaps. Here is why you need all four.
2026-02-13
You know your organization needs AI governance. Your CISO is skeptical. Here is the internal champion playbook -- what CISOs care about, how to frame the conversation, and how to build the case that gets budget approved.
2026-02-12
Binary allow/deny decisions are insufficient for AI agent governance. Real-world actions exist on a risk continuum. Here is how Intended calculates risk dynamically using eight dimensions of context.
2026-02-11
For compliance engineers managing SOC 2, ISO 27001, or industry-specific frameworks, Intended provides automated evidence collection, chain verification, and auditor-ready exports. Here is how to map Intended to your compliance controls.
2026-02-10
Not every organization can send AI governance data to the cloud. Defense, financial services, healthcare, and critical infrastructure often require air-gapped or on-premise deployment. Here is how Intended approaches each deployment model — managed cloud is generally available today; on-premise and air-gapped are a gated offering, available on request for qualifying organizations.
2026-02-09
GitHub, Jira, Salesforce, and ServiceNow all send webhooks in different formats. Intended normalizes them into a single unified intent format so your policies work across every system without system-specific rules.
2026-02-08
A million governance decisions per month is enterprise-grade throughput — and every millisecond of latency and every bottleneck in the pipeline matters at that volume. Here is how Intended's architecture is designed to scale horizontally to handle it.
2026-02-07
We went through SOC 2 Type II preparation ourselves. Here is a transparent account of what was harder than expected, what was easier, and what we would do differently if we started over.
2026-02-06
Deciding what to open-source and what to keep commercial is one of the hardest strategic decisions a platform company makes. Here is how we made that decision, and what we learned from Databricks, HashiCorp, and the broader industry.
2026-02-05
AI governance has its own vocabulary. Intent, authority token, domain pack, Open Intent, fail-closed, risk score, evidence bundle -- here are 40-plus terms defined clearly and precisely so everyone speaks the same language.
2026-02-04
The AI governance landscape is shifting fast. The EU AI Act is entering enforcement, autonomous agents are proliferating, and multi-agent systems are going production. Here is where the industry is headed and what it means for governance.
2026-02-03
A full technical architecture post for CTOs and senior engineers. Every component of the Intended platform explained -- from the intent compiler to the hash-chained audit ledger, with data flows, scaling characteristics, and design rationale.
2026-02-02
When an AI agent does something wrong -- an unauthorized action, a misconfiguration, a data leak -- you need a playbook. Detection, containment, investigation, and remediation for AI agent incidents.
2026-02-01
Version-controlled policies, Git-based review workflows, and CI/CD for governance. Here is how to treat your AI governance policies with the same rigor as your application code.
2026-01-31
The Open Intent taxonomy classifies AI agent actions into 29 domains and 173 categories. Here are the design principles that guided its creation and why those principles matter for governance at scale.
2026-01-30
LangChain, PydanticAI, CrewAI, OpenAI Agents SDK -- none of them have built-in governance. They all provide tool calling without authority. Here is why every AI framework needs an authority layer, and why that layer should be external.
2026-01-29
The model for building custom connectors on Intended — and the honest status today: the connector toolkit is internal (private preview), and the public build-your-own API is on the roadmap.
2026-01-28
Governance without observability is governance in the dark. Here is how to monitor AI agent decisions in real time -- metrics, dashboards, alerts, and the signals that matter most.
2026-01-27
Where your governance data lives matters more than ever. GDPR, data sovereignty laws, and enterprise requirements demand control over data location. Here is how Intended handles multi-region deployment and data residency.
2026-01-26
Building the ROI case for AI agent governance. Risk reduction, time savings, compliance value, and the cost of doing nothing. A framework for executive presentations.
2026-01-25
A roundup of what we shipped in early 2026. MCP Gateway for model context protocol governance, Python SDK, Kubernetes admission controller, new pricing tiers, and 15 new blog posts for the community.
2026-01-24
API keys are the credentials your AI agents use to interact with Intended. Rotation, scoping, grace periods, and monitoring. Here are the best practices for managing API keys in a governance-critical system.
2026-01-23
Manual review of AI agent actions does not scale. At 50 agents making 500 decisions a day, you need a team just to review. Automated governance replaces manual review without sacrificing control.
2026-01-22
SaaS platforms deploying AI agents face unique governance challenges. Per-tenant policies, data isolation, usage metering, and cross-tenant security. Here is how to implement AI governance in a multi-tenant architecture.
2026-01-21
A technical deep-dive into hash-chained audit trails. SHA-256 chains, serializable transactions, tamper detection, and why traditional logging is insufficient for AI governance compliance.
2026-01-20
The quickest possible path from zero to governed AI agent. Sign up, install the SDK, submit your first intent, and see the governance decision. Five minutes, no infrastructure required.