The problem

AI cannot act safely on stale or unattributed information.

The moment software moves from answering questions to taking actions, the question changes from "what does the model know" to "what evidence supported this decision".

Models carry outdated knowledge.
Training data has a cutoff. A system acting on the world needs to know what is true now, not what was true when the model was built.
Search results are not authoritative.
Ranked links are not evidence. Machine decisions need information tied to a named source, not to whatever surfaced first.
Machine decisions need auditable evidence.
When an automated action is questioned later, the inputs it relied on must still exist, unchanged, with their timestamps and origins intact.
Enterprises and governments need provenance and accountability.
Regulated organisations cannot adopt automated decision-making they cannot explain, attribute and review.

Platform

What the evidence layer provides.

Seven capabilities, one purpose: every machine decision can show its working.

Live verified signals

Current readings from external sources, served with integrity checks applied at response time, so automated systems act on what is happening rather than on what was cached.

Source provenance

Every datapoint names its publisher and licence. Attribution travels with the data.

Freshness metadata

Age and observation time on every response, so consumers can tell live from stale.

Evidence receipts

Signed records preserving what a decision relied on, verifiable independently later.

Multi-source reconciliation

Independent sources compared side by side, disagreements surfaced rather than hidden.

Audit trails

Append-only archives of served evidence, retained for later review and discovery.

Human oversight

Evidence laid out in full so a person can review high-consequence decisions before acting on them.

Products

The first product of the evidence layer.

DynamicFeedby Fluxdyne

Live, verifiable real-world data for AI agents, enterprise systems and government workflows. Every response is signed, source-attributed and independently verifiable.

Visit DynamicFeed View MCP Read Documentation

REST and MCP, keyless to startAgents connect over standard protocols without an onboarding call.

Signed responsesEd25519 signatures with published keys and open-source verifiers.

Deterministic and machine-firstStable schemas built to be consumed by software, not screens.

Developers

Your first signed response in under a minute.

Reads are keyless. One REST endpoint, one MCP server, the same signed envelope everywhere. Full reference at dynamicfeed.ai/connect.

REST

curl -s https://dynamicfeed.ai/v1/batch \
  -H "Content-Type: application/json" \
  -d '{"calls":[{"tool":"earthquakes",
       "args":{"min_magnitude":5,"limit":3}}]}'

Signed Ed25519 envelope on every response. No key, no signup for reads.

MCP

{
  "mcpServers": {
    "dynamicfeed": {
      "url": "https://dynamicfeed.ai/mcp"
    }
  }
}

Or run it locally: npx dynamicfeed-mcp. Works with any MCP client.

Live proof

Fetch a real response from the live API, right from this page.

Receipts

Verify a receipt in three steps.

Verification runs in your browser against published keys. Nothing is uploaded. A valid receipt proves the integrity and attribution of a snapshot, not the truth of the underlying facts.

Get a signed artifact. Any DynamicFeed response, Decision Receipt, or one of the published sample fixtures.
Open the verifier. Paste the artifact, or load a live sample with one click.
Read the verdict honestly. Signature, key lifecycle and freshness are graded separately, and the boundary of what a pass proves is stated on the page.

Open the verifier What receipts prove

Solutions

Where auditable evidence matters most.

Enterprise AI

Give internal AI systems inputs you can defend.

Enterprise copilots and decision systems inherit the liability of their inputs. Fluxdyne attaches source, licence, freshness and verification state to the data those systems consume, so risk and audit teams can trace any output back to what informed it.

Government

Automated decisions that survive review.

Public-sector automation is held to a higher standard of explainability. Evidence receipts preserve what a workflow relied on at the moment it acted, supporting administrative review, records obligations and public accountability.

AI Agents

Ground agent actions in verifiable data.

Agents act at machine speed on information they did not create. Fluxdyne gives them a data layer that is current, attributed and cryptographically signed, over the same protocols agents already speak.

Autonomous Systems

A record of what the machine knew.

Robots, fleets and industrial systems act on external conditions. When an action is questioned, the evidence of those conditions, as observed at decision time, should still exist, timestamped and intact.

Current status

Private research and development.

Fluxdyne is currently in private research and development. External demonstrations and design-pilot discussions will reopen following company formation.

Please do not submit confidential, personal, or commercially sensitive information.

Sovereign capability

Australian-developed infrastructure for trusted AI.

Fluxdyne is an Australian company building evidence infrastructure with global reach.

Developed and operated from Australia, serving a global market.
Sovereign engineering capability in a layer of the AI stack where provenance and accountability are becoming procurement requirements.
Designed for enterprise and government readiness, with auditability, human review paths and source attribution built in from the start.
Export-oriented infrastructure software in a category with worldwide demand.
A responsible-AI posture grounded in evidence: our systems record what informed a decision rather than certifying that the decision was correct.

Company

Machines are starting to act. We keep the evidence.

Fluxdyne is building infrastructure that helps machines understand what is current, where it came from and what evidence supported a decision. We believe the next layer of the AI stack is not a bigger model. It is the layer that lets automated decisions be examined, attributed and answered for.

Evidence infrastructure for machine decisions.