Geordie AI
by Geordie AI Ltd.
Discover, govern and cost-account every AI agent running across your enterprise
Geordie is an AI agent security and governance platform that discovers every agent running across an enterprise's cloud, code, endpoint and SaaS estate, maps what each one can reach, keeps an auditable record of what it actually does, and now attributes AI spend down to the individual agent that caused it. It is built for multi-vendor agent environments rather than a single model provider.
Geordie is a security and governance platform for AI agents, built for enterprises running agents from several vendors at once. It works from an endpoint-based vantage point across cloud, code, endpoint and SaaS environments rather than from a gateway or router, which the company argues is what lets it see the full picture of what an agent does and what it costs. Four capabilities sit on that foundation. Agent discovery and posture management finds every agent in the estate and maps its permissions, tool connections, knowledge sources and system prompts, updating continuously — at biotech customer Owkin it surfaced three times more agents than the company knew it had. Behavioural observability keeps an auditable record of prompts, plans, responses and tool invocations, with behavioural baselining and anomaly detection. Risk intelligence applies proprietary threat modelling and maps findings to OWASP, NIST, ISO 42001 and the EU AI Act. Beam delivers real-time, agent-level remediation without gateway latency, using context engineering to shape and constrain agent behaviour, offering either deterministic interventions or human-oversight advisories. On 9 September 2026 Geordie added Cost Intelligence, which attributes AI spend to the specific agent, task and owner that generated it — drilling from an organisation-wide total down to platform, team, user, model and individual workflow, tracking per-agent cache efficiency as its own metric, and flagging agents stuck in tool loops or routed to models far larger than the job needs. Founded in 2025 by former Darktrace and Snyk leaders, Geordie has raised $36.5 million and runs roughly 30 customer environments including AlphaSense, Owkin, OakNorth and A+E Global Media.
The CISO or platform-security lead at an enterprise already running agents from more than one model vendor, who cannot answer how many agents exist, what each can reach, or which one is burning the AI budget.
One telemetry stream that inventories every agent, records what it did, constrains it in real time, and attributes its spend to a named owner.
At a Glance
- Category
- Governance & Security
- Pricing
- Contact for pricing
- Target Market
- CISOs, CIOs, CTOs, Platform Engineering Leads, FinOps Teams
- Deployment
- Cloud-first
- Founded
- 2025
- Headquarters
- London, United Kingdom
- Team Size
- 11-50
- Customers
- ~30 customer environments as of May 2026
Key Features
- ✓Agent discovery and posture management
Finds agents nobody registered and maps their permissions, tool connections, knowledge sources and system prompts continuously, not just at onboarding.
- ✓Behavioural observability
Keeps an auditable record of prompts, plans, responses and tool invocations, with baselining so anomalous agent behaviour actually stands out.
- ✓Risk intelligence mapped to frameworks
Proprietary threat modelling with findings mapped to OWASP, NIST, ISO 42001 and the EU AI Act, so audit evidence falls out of the tool.
- ✓Beam real-time remediation
Agent-level controls applied without gateway latency, using context engineering to constrain behaviour deterministically or route to a human for oversight.
- ✓Cost Intelligence attribution
Drills AI spend from an organisation-wide total down to platform, team, user, model, individual agent and workflow.
- ✓Per-agent efficiency and waste detection
Tracks caching efficiency per agent and flags infinite tool loops, oversized model routing and sessions burning tokens with little output.
- ✓Endpoint-based multi-vendor coverage
Sits close to where agents actually run across cloud, code, endpoint and SaaS, so it is not tied to one model provider.
Use Cases
- •Shadow agent discovery
Owkin found three times more agents than it knew about, across hundreds of agents and more than 50 petabytes of data.
- •AI cost control
One retail customer was losing roughly $1 million a month routing tasks to models more expensive than the work required.
- •Runaway agent containment
At a financial institution, an agent stuck in a tool loop consumed half the monthly AI budget in a single day.
- •EU AI Act and ISO 42001 evidence
Risk findings map directly to the frameworks auditors ask about, replacing a manual control-mapping exercise each cycle.
- •Agent incident investigation
The behavioural record reconstructs what an agent planned, which tools it called and what it returned when something goes wrong.
Ideal For
Best For
- ✓CISOs and platform teams governing agents from multiple model vendors rather than one provider
- ✓Enterprises that genuinely do not know how many AI agents are already running in their estate
- ✓FinOps and platform owners who need AI spend attributed to an owning team rather than a token total
- ✓Regulated organisations needing EU AI Act, ISO 42001, NIST or OWASP evidence for agent deployments
- ✓Security teams wanting agent-level intervention without putting a gateway in the request path
Not Ideal For
- ✗Organisations standardised on a single vendor's stack, where Microsoft Agent 365 or ServiceNow AI Control Tower may already cover the need — Geordie's own CEO names incumbent distribution as his main competitive risk
- ✗Buyers who need published list pricing to build a business case, since Geordie discloses none anywhere
- ✗Teams not yet running agents in production, where discovery and behavioural baselining have nothing to work on
- ✗Anyone expecting a lightweight SaaS connector, since coverage requires endpoint-close instrumentation across cloud, code and endpoints
Deployment
Market & Ratings
~30 customer environments as of May 2026
Market Analysis
Pros
- ✓Founding team drawn from Darktrace and Snyk, with an unusually strong award record (RSA Conference Innovation Sandbox winner 2026, Gartner Cool Vendor 2026, Black Hat Europe Startup Spotlight 2025)
- ✓Endpoint-based approach avoids inserting a latency-adding gateway into the agent request path
- ✓Security, governance and cost sit on one telemetry stream rather than requiring three separate products
- ✓SOC 2 Type II, SOC 3, ISO/IEC 27001 and GDPR, with a public SafeBase trust centre carrying pentest and SOC 2 reports
- ✓Customer outcomes are published with names attached (Owkin, AlphaSense, Wood Partners) rather than anonymous logos
Cons
- ✗No published pricing of any kind, so there is no way to scope a budget without entering a sales cycle
- ✗Competing head-on with Microsoft Agent 365, ServiceNow AI Control Tower and OpenAI Frontier at roughly 37 employees — the CEO himself names incumbent distribution advantage as the primary threat to the company
- ✗Under two years old with effectively no independent review base: no usable G2, Capterra or TrustRadius ratings and no Hacker News discussion, so every performance claim traces back to the vendor or to press coverage of the vendor
- ✗Deployed across roughly 30 customer environments, a small base from which to judge reliability at enterprise scale
- ✗Endpoint-close instrumentation across cloud, code, endpoint and SaaS is a rollout project rather than a connector you switch on
- ✗Reported Series A post-money valuation differs between sources ($155M in one, roughly $180M in another), a reminder that the funding detail is press-reported rather than filed
Pricing
Enterprise
Contact for pricing
- ✓Agent discovery and posture management
- ✓Behavioural observability
- ✓Risk intelligence mapped to OWASP/NIST/ISO 42001/EU AI Act
- ✓Beam real-time remediation
- ✓Cost Intelligence
Geordie publishes no list pricing or plan tiers anywhere, including across its Series A and product-launch coverage, so every evaluation goes through enterprise sales. Cost Intelligence is described as a capability of the platform rather than a separately priced module.
Security & Compliance
Sources
This page was written from 6 sources, 4 on domains other than geordie.ai.
- 1.geordie.ai — geordie.aivendor
- 2.geordie.ai — trustvendor
- 3.fortune.com — geordie security governance ai agents
- 4.siliconangle.com — geordie launches cost intelligence to tie ai spending to age
- 5.dealroom.co — 129954 geordie ai raises 30m series a led by balderton for a
- 6.techfundingnews.com — geordie raises 30m series a led by balderton to become the s
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
Broadcom AgentMinder
Authorize every AI agent action before it happens, not audit it afterwards
CrowdStrike Falcon Guardian
AI detection and response that finds shadow AI agents on the endpoint and blocks the unapproved ones
Tenable CyberAgents Exchange
A free, vendor-neutral registry of security AI agents, skills, MCP servers and playbooks
Kosmoy
A self-hosted AI control plane: inventory, gateway, observability and agent sandboxing