Pathiqa

Governed infrastructure for enterprise AI integration

Everything your agents need to know and do: in context, governed, and on record.

Enterprise data was never built for AI. It is fragmented across systems that share no language, guarded by permissions built for humans, leaving agents under-exposed to the context they need or over-exposed to everything else. Pathiqa computes an ontology across every source, scopes each agent to the context and actions it needs, and records everything it reads or does on an immutable ledger.

Ontology · Scopes · Ledger · Views

What we're solving

An agent is only as good as the context it uses. Existing enterprise data infrastructure cannot give AI systems the context they need to execute autonomous cross-functional workflows.

Data fragmentation slows execution and degrades agentic decision-making by forcing agents to act on incomplete and inconsistent context. Existing enterprise access systems built on static, role-based, and application-bound permissions are ill-suited to AI systems requiring dynamic, task-based permissions, leaving agents under-exposed to the context they need or over-exposed to access they don’t.

An agent is only as good as the record it leaves. Enterprise agents acting on numerous data sources, platforms, and tools leave a fragmented record of activity. It spreads across third-party systems, each with its own owner, format, and policies. With no single account of what an agent did and why, an enterprise cannot securely govern its decision-making, which deters AI deployment at scale.

The architecture

One architecture.
Five layers of value. Three components.

// 01

The ontology

Connects · Insulates

Resolves the enterprise’s data by meaning and persists independently of the systems it describes.

// 02

The ledger

Governs · Learns

Records every declared intent, access, action, and outcome as a signed, immutable entry.

// 03

The key

Exposes · Carries reputation

The unit to which scope, access rights, and reputation attach, internally and across boundaries.

Five layers of value

Pathiqa's approach creates five layers of value for an enterprise.

// 01

Connect

// Ontology

Agents work with the context they need.

An agent that cannot resolve a customer across CRM, billing, support and product fails the moment it is deployed. Pathiqa’s ontology resolves meaning across the existing estate without consolidation. Requests from agents are given only authorised, task-relevant context, so agents produce better answers and outcomes. Our intelligent routing and permission-aware caching cut retrieval cost and load. The agent receives a coherent semantic surface that adapts as schemas evolve.

ForCIO · Head of AI · Business sponsor · CFOAligns toEU AI Act Art. 14 · NIST AI RMF Map
// 02

Govern

// Ledger

Every action is declared, recorded and, where supported, reversibly controlled.

Before an agent can access data, Pathiqa requires it to declare its intent, scope, and boundaries. This declaration becomes an enforceable contract. Every action is written to a structured, immutable ledger: intent, data accessed, model used, outcome produced. When something goes wrong and the underlying system supports rollback or a compensating action, versioned rollback reverses a single agent’s impact without affecting other agents, users, or operations.

ForCISO · General Counsel · CRO · ComplianceAligns toEU AI Act Art. 12 · ISO 42001 · DORA
// 03

Learn

// Ledger

The ledger compounds into a record of how the enterprise actually works.

Every agent decision is captured with intent, scope, model, latency, cost, and outcome. Over time this accumulates into operational awareness an enterprise has not had before: where work happens and what it costs, which agents and models deliver, what to automate further or retire. It is also the foundation for cost governance. Once the enterprise knows what a task should cost, budgets attach to keys, scopes, and outcomes.

ForCDO · COO · CEO · CFOAligns toEU AI Act bias monitoring · algorithmic accountability
// 04

Expose

// Key

Enterprise data, exposed to the outside on commercial terms.

Pathiqa’s ‘key’ architecture gives an enterprise a new, and safe, way to share their data with their customers or the public. The key is the unit to which reputation, scope, and access rights attach. Customer-, partner-, and public-facing keys each receive carefully scoped access, with every action recorded and every contract violation triggering automatic revocation. Banks let counterparty agents query positions; insurers let broker agents quote against policy data. Pathiqa moves from cost centre to revenue platform.

ForCCO · Partnerships · BD · ProductAligns toGDPR Art. 6 & 7 · Open Banking
// 05

Insulate

// Ontology

The enterprise stays in control of its data and its history, independent of any vendor.

Pathiqa is deliberately on the enterprise’s side. Adopting it gives an enterprise the freedom to change its software and AI vendors whenever it wants, avoiding expensive lock-in and staying in full control of its data. The enterprise, rather than its software providers, owns its ontology and ledger. Schemas can change, vendors can be swapped, models can be replaced.

ForBoard · Audit Committee · CRO · CIOAligns toDORA Art. 28–30 · EU AI Act audit continuity
Posture

Pathiqa is vendor-neutral on both sides. Above it, agents may be foundation models, bespoke reasoning systems, or human-in-the-loop copilots. Beneath it, systems of record may be core line-of-business systems, document stores, or warehouses. The runtime speaks to both, so neither has to speak to the other directly.

The runtime is a layer for metadata and decisions. Raw data remains in its system of record unless an authorised task explicitly requires transmission. Pathiqa mediates what is released, to whom, for what purpose and under which policy.

Views

A View is a living answer.

Views are Pathiqa’s product surface: a shareable, continuously updating window onto your enterprise. A View draws on every data estate at once, sets the numbers beside an agent’s read of them, surfaces what you would not have known to ask, and answers follow-up questions in place, within your permissions.

Continuously updating

Not a static report. A View recomputes as the data and the agents move, so what you open is always current.

Across every estate

A View spans CRM, billing, support, and warehouses at once, with no consolidation into a new dashboard stack.

Numbers and judgement

A deterministic count from a database sits beside an agent’s assessment of what it means.

Proactive

Pathiqa surfaces what you would not have known to request, so you never re-run the same query.

Conversational

The recipient can question a View in place, within its scope and permissions.

Shareable

Internally and externally, scoped by key, so access and reputation travel with the View.

Agent Stand-Up

Treasury Ops

Live · updated 2 min ago
142Invoices resolved
3Exceptions
68%Spend vs budget
// agent · operational assessment

Cash position is healthy. Two counterparties, ACME and Northwind, are slow to confirm, so I’ve queued ACME for manual review. One refund exceeds my policy limit and needs a human decision.

confidence: medium
⚑ 1 refund over policy limit, awaiting your decision

// first View · the Agent Stand-Up: a daily brief from your agents on what they did, what they are unsure about, and what needs a human.

Suggestion & control

Pathiqa proposes the Views, and guards what leaves.

Because Pathiqa understands your data by meaning, it can both recommend the Views worth having and reason about what each one should reveal, and to whom.

Suggested Views

Pathiqa reads your data and its semantic context and proposes the Views worth having, including the ones you would not have thought to build. As the estate changes, new questions surface on their own.

Counterparty concentration

// exposure to Northwind up 40% this quarter

+ Add

Refunds above policy

// 3 pending, on an upward trend

+ Add

Idle entitlements

// 12 keys unused for 30 days

+ Add

Filters and lenses

Before a View leaves the building, Pathiqa evaluates what can be shared under your disclosure policies. Controllable filters and per-recipient lenses hold back sensitive or strategic information, so an external View never reveals more than you intend.

Share · Q3 Treasury Stand-Up

Recipient: Northwind Capital (external)

InternalPartnerPublic
Margin by deskWithheld · strategic
Pipeline forecastWithheld · strategic
Account-holder detailsRedacted · policy
// pathiqa · advises

Sharing margin-by-desk would expose your pricing strategy to a counterparty. The Partner lens holds it back.

Ready to use

Integrate an agent in a few lines.

Pathiqa ships idiomatic clients for every stack. Point an agent at a question and get an answer grounded in your governed data, with every access written to the immutable ledger. The same call, in your language.

from pathiqa import Pathiqa

client = Pathiqa()

response = client.ask(
    agent="ag_1e6477e5812e424ef9d087effa4670ec",
    question="How have time-of-day and day-of-week purchasing patterns changed since the launch of our new ad campaign in Tokyo over the last 3 weeks?",
)

print(response.output_text)
import { Pathiqa } from "pathiqa";

const client = new Pathiqa();

const response = await client.ask({
  agent: "ag_1e6477e5812e424ef9d087effa4670ec",
  question: "How have time-of-day and day-of-week purchasing patterns changed since the launch of our new ad campaign in Tokyo over the last 3 weeks?",
});

console.log(response.outputText);
import ai.pathiqa.Pathiqa;
import ai.pathiqa.models.AskParams;
import ai.pathiqa.models.AskResponse;

Pathiqa client = new Pathiqa();

AskResponse response = client.ask(AskParams.builder()
    .agent("ag_1e6477e5812e424ef9d087effa4670ec")
    .question("How have time-of-day and day-of-week purchasing patterns changed since the launch of our new ad campaign in Tokyo over the last 3 weeks?")
    .build());

System.out.println(response.outputText());
package main

import (
    "context"
    "fmt"

    "github.com/pathiqa/pathiqa-go"
)

func main() {
    client := pathiqa.NewClient()

    response, err := client.Ask(context.Background(), pathiqa.AskParams{
        Agent:    "ag_1e6477e5812e424ef9d087effa4670ec",
        Question: "How have time-of-day and day-of-week purchasing patterns changed since the launch of our new ad campaign in Tokyo over the last 3 weeks?",
    })
    if err != nil {
        panic(err)
    }

    fmt.Println(response.OutputText)
}
use pathiqa::Pathiqa;

let client = Pathiqa::new();

let response = client
    .ask()
    .agent("ag_1e6477e5812e424ef9d087effa4670ec")
    .question("How have time-of-day and day-of-week purchasing patterns changed since the launch of our new ad campaign in Tokyo over the last 3 weeks?")
    .send()
    .await?;

println!("{}", response.output_text);
require "pathiqa"

client = Pathiqa::Client.new

response = client.ask(
  agent: "ag_1e6477e5812e424ef9d087effa4670ec",
  question: "How have time-of-day and day-of-week purchasing patterns changed since the launch of our new ad campaign in Tokyo over the last 3 weeks?"
)

puts response.output_text
use Pathiqa\Client;

$client = new Client();

$response = $client->ask(
    agent: 'ag_1e6477e5812e424ef9d087effa4670ec',
    question: 'How have time-of-day and day-of-week purchasing patterns changed since the launch of our new ad campaign in Tokyo over the last 3 weeks?',
);

echo $response->outputText;
using Pathiqa;

var client = new PathiqaClient();

var response = await client.AskAsync(
    agent: "ag_1e6477e5812e424ef9d087effa4670ec",
    question: "How have time-of-day and day-of-week purchasing patterns changed since the launch of our new ad campaign in Tokyo over the last 3 weeks?"
);

Console.WriteLine(response.OutputText);
The DSI stack

Two products, one platform.

Pathiqa governs how agents reach an enterprise’s data. Exostrate governs the agents themselves, packaging, isolating, budgeting, and auditing the runtime. Together they close the loop from the record an agent leaves to the runtime that produced it, grounded in the same research programme.

Deployment
Hosted
A managed instance operated by DSI. The fastest route into a working pilot.
Dedicated
A single-tenant instance in a dedicated environment. For organisations with stricter isolation requirements.
On-premise
Installed inside the organisation’s own environment. For sovereign and high-sensitivity workloads.
Access

Pathiqa is in early deployment with a select group of design partners. New engagements begin with a short briefing.

Request access