Operational truth layer
Know what your automation
actually did.
Not what it reported. Mindplane sits inside the action and issues a receipt for every run: the path it declared, the path it observed, and the diff between them.
A verifiable trail across tools and environments, keyed by correlation ID. Nothing is trusted because it was claimed; only what was observed counts as true.
Running automation or agents in production? Request a pilot →
What Mindplane is
- WhatThe operational truth layer: it sits between intent and execution and issues a tamper-evident receipt for every automated action: declared path, observed path, and the diff.
- WhoBuilt for SRE / platform, ops, and AI / ML teams running automation or agents in production.
- ProblemWhen a run goes wrong, teams reconstruct what happened from logs that record what was reported, not what was observed. The two disagree exactly when it matters.
- OutcomeReceipts you can query by correlation ID (intent, context, validators, observed outcome) so "who changed what?" has one answer, not five.
Where Mindplane fits
Where it sits
- →Inside the action, between intent and execution
- →Captures the decision at the source, not after the fact
- →One receipt per run, keyed by correlation ID
- ✗Not a policy engine (yet)
- ✗Not an after-the-fact log aggregator
Why it reduces pain
- ✓When a run fails, the receipt shows which step diverged
- ✓One account of "what happened," not several
- ✓Observed state separated from inferred state
- ✓Drift surfaced as a diff, not a guess
- ✓A single trail that holds across tools
Three places automation lies by accident
- AuthorityWho was actually allowed to act?
- DataWas that observed state, or fabricated state?
- ExecutionDid it follow the path it declared?
Mindplane receipts the difference.
What's inside a receipt
Correlation ID
Every action carries one ID across tools and environments. cog corr get reassembles the whole run from it.
Conformance
The path the action declared, checked against the path actually observed. The diff is the product.
Tamper-evident ledger
Entries can't be edited or deleted. Denied and drifted runs are recorded too, not just the clean ones.
Observed beats inferred
Only observed state can satisfy an invariant. Inferred state can inform a receipt; it can never authorise.
Multi-tenant by default
Enforced at the API boundary, not in application logic.
Queryable
Every run is queryable and explainable by correlation ID. Reconstruct any run, clean or failed.
The team
Based in Belfast & remote
Scott Alexander
Founder & CEO
SRE leader, 10+ years building scalable infrastructure
Building operational truth systems under production pressure
Focus: observed truth, auditability, operational decision trails
Jay Kinder
Co-Founder & CTO
Principal SRE focused on automation and platform engineering
Designing reliable execution paths and production-safe automation
Focus: implementation rigor, systems design, production delivery
Live today
V1 is a multi-tenant operational truth layer, running in production and ingesting real events from Datadog, Grafana, and custom webhooks.
- Multi-tenant receipt ledger
- Declared-versus-observed run diff
- Correlation ID lookup (
cog corr get) - Run receipt and Decision receipt
- Observe-only mode, before anything blocks
Live today — receipts, correlation, declared-vs-observed diff, observe-only mode
Building next — enforcement mode, reconciliation receipts, policy enforcement boundaries
Roadmap — autonomous enforcement, full authority resolution, Assembly, ActionStore
Built for observe-only pilots
Mindplane is designed to start safely: observe first, record receipts, compare declared work with observed evidence, and show what automation would have done before it is trusted to act.
- ✓Best fit: teams running deployment, incident, infrastructure, or agentic automation
- ✓First pilot: observe-only, no production blocking
- ✓Success measure: fewer blind spots between what automation reported and what actually happened
Contact
