LLM-Native Browser

Agentic browsing binds observation, action, authority, and evidence into controlled execution—not merely page viewing.

Engineer DOM, accessibility, and visual state with session isolation, tool authority, approvals, and reproducible execution records.

§ 01

Problem definition

The operating conditions that justify LLM-Native Browser

On dynamic pages, divergence between model observation and click target can record incorrect actions as successful.

A browser mixing cookies, credentials, and external input becomes a boundary for prompt injection and authority misuse.

  • Human web workflows without APIs must be automated audibly
  • Pre-action approval and result evidence must be preserved at each step
  • A stable official API performs the same work
  • Credential isolation and distrust of external content cannot be enforced
PLATE 01

LLM-Native Browser: system plate

  1. 01

    State 1

    Capture page observation as synchronized DOM, accessibility, visual, and network evidence.

  2. 02

    State 2

    Separate read, input, submit, and download authority with approval before irreversible actions.

  3. 03

    State 3

    Separate instructions from external content from user and policy instructions in the execution layer.

  4. 04

    State 4

    Retain pre/post-action state and failure causes in an event log and replayable run bundle.

FEEDBACKFailed acceptance returns evidence to the first controlled stage: Task success and incorrect-action rate.

A decision and validation view for LLM-Native Browser; labels describe architecture, not a measured deployment result.
  1. The workflow begins with Capture page observation as synchronized DOM, accessibility, visual, and network evidence..
  2. It reaches an acceptance decision through Task success and incorrect-action rate.

§ 03

Design method

Fix the boundary and acceptance criteria before implementation.

Agentic browsing binds observation, action, authority, and evidence into controlled execution—not merely page viewing.

  1. 01

    Stage 1

    Capture page observation as synchronized DOM, accessibility, visual, and network evidence.

    Review artifact 1
  2. 02

    Stage 2

    Separate read, input, submit, and download authority with approval before irreversible actions.

    Review artifact 2
  3. 03

    Stage 3

    Separate instructions from external content from user and policy instructions in the execution layer.

    Review artifact 3
  4. 04

    Stage 4

    Retain pre/post-action state and failure causes in an event log and replayable run bundle.

    Review artifact 4

§ 04

Application scenarios

Hypothetical workloads make the applicability boundary concrete.

Hypothetical application scenario

Human web workflows without APIs must be automated audibly

On dynamic pages, divergence between model observation and click target can record incorrect actions as successful.

APPROACH
Capture page observation as synchronized DOM, accessibility, visual, and network evidence.
BOUNDARY
Demo success is not generalized to reliability on changing live websites.
Hypothetical application scenario

Pre-action approval and result evidence must be preserved at each step

A browser mixing cookies, credentials, and external input becomes a boundary for prompt injection and authority misuse.

APPROACH
Separate read, input, submit, and download authority with approval before irreversible actions.
BOUNDARY
Browser authority must not exceed user authority.

§ 05

Design choices

Review gains and costs in the same table.

DecisionGainCostWatch
Human web workflows without APIs must be automated audiblyCapture page observation as synchronized DOM, accessibility, visual, and network evidence.Demo success is not generalized to reliability on changing live websites.Task success and incorrect-action rate
Pre-action approval and result evidence must be preserved at each stepSeparate read, input, submit, and download authority with approval before irreversible actions.Browser authority must not exceed user authority.Approval and authority violation detection
PLATE 02

LLM-Native Browser: system plate

Client boundaryHuman web workflows without APIs must be automated audibly
  • On dynamic pages, divergence between model observation and click target can record incorrect actions as successful.
  • Demo success is not generalized to reliability on changing live websites.
PattyCapture page observation as synchronized DOM, accessibility, visual, and network evidence.
  • Separate read, input, submit, and download authority with approval before irreversible actions.
  • Task success and incorrect-action rate
AcceptanceApproval and authority violation detection
  • Run reproducibility and evidence completeness
  • Browser authority must not exceed user authority.
A decision and validation view for LLM-Native Browser; labels describe architecture, not a measured deployment result.
  1. The workflow begins with Capture page observation as synchronized DOM, accessibility, visual, and network evidence..
  2. It reaches an acceptance decision through Task success and incorrect-action rate.

§ 07

Validation plan

Agree on measurement conditions before publishing a result.

MeasureMethodPass conditionCaveat
Task success and incorrect-action rateCapture page observation as synchronized DOM, accessibility, visual, and network evidence.Repeated runs satisfy the acceptance threshold agreed during discoveryDemo success is not generalized to reliability on changing live websites.
Approval and authority violation detectionSeparate read, input, submit, and download authority with approval before irreversible actions.Repeated runs satisfy the acceptance threshold agreed during discovery
Run reproducibility and evidence completenessSeparate instructions from external content from user and policy instructions in the execution layer.Repeated runs satisfy the acceptance threshold agreed during discovery

§ 08

Constraints and failure conditions

Conditions for not applying the capability are part of the design.

A stable official API performs the same work

Demo success is not generalized to reliability on changing live websites.

Credential isolation and distrust of external content cannot be enforced

Browser authority must not exceed user authority.

§ 10

Durable deliverables

Artifacts remain with the operating organization after the engagement.

LLM-Native Browser decision record
Agentic browsing binds observation, action, authority, and evidence into controlled execution—not merely page viewing.Client-owned · Patty-reviewed
Validation harness and acceptance criteria
Task success and incorrect-action rate · Approval and authority violation detection · Run reproducibility and evidence completenessJointly maintained
Operations and recovery runbook
Demo success is not generalized to reliability on changing live websites. · Browser authority must not exceed user authority.Operating-team owned

§ 11

Terminology

Use shared terms with explicit operating meaning.

LLM-Native Browser
Engineer DOM, accessibility, and visual state with session isolation, tool authority, approvals, and reproducible execution records.
Acceptance criterion
Task success and incorrect-action rate
Operating boundary
Demo success is not generalized to reliability on changing live websites.

REFERENCES

References and primary material

  1. W3C WebDriver

    Primary material for the method and terminology.

  2. OWASP LLM Prompt Injection

    Primary material for the method and terminology.

Begin by determining whether LLM-Native Browser is the justified next step.

We define scope and validation against representative work, data and infrastructure boundaries, and explicit failure conditions.

Request a technical review