Skip to content
SpecStudio
Start free

Specs for browser AI agents

Hand your browser agent a spec, not a wish.

SpecStudio turns a one-line request into a task specification an agent can follow and you can audit: scope, allowed domains, steps with checks, and the approval checkpoints a deterministic policy engine requires.

Compile your first spec free See pricing
  • Free plan with no card. Sign in with Google in one click.
  • A rules engine, not the model, decides what needs your approval.
  • Never runs the task and never stores credentials.
SpecStudio: an informal request on the left, a spec assurance score and validation results on the right
SpecStudio: an informal request on the left, a spec assurance score and validation results on the right. Real product screen.

Why agents go sideways

A vague request gives you an agent that is almost right.

“Find the cheapest refundable flight and book it” leaves out the budget, the dates, which sites are allowed and the moment a human must say yes. The agent fills the gaps on its own, then hits a CAPTCHA, a login wall or an injected instruction on a web page. A spec closes those gaps before the first click.

66%

of developers name “AI solutions that are almost right, but not quite” as their top frustration with AI tools.

Source: Stack Overflow Developer Survey 2025

From a rough request to an approved spec in four steps

  1. 01

    Write it the way you would say it

    Paste the request, then add allowed domains, constraints and what success looks like. Start from a template if you like.

  2. 02

    Compile

    The model drafts the spec. Your request stays the only trusted instruction; quoted web content is treated as data.

  3. 03

    Policy pass

    A deterministic engine scores the spec, forces checkpoints on purchases, sends and deletes, and flags missing information.

  4. 04

    Export and run

    Copy it as Markdown or JSON, send it to the workbench, or export AGENTS.md, CLAUDE.md and Cursor rules.

Everything the agent needs, nothing it can misread

Each spec separates trusted intent from untrusted content, references page elements by their visible labels, and states how every step is verified.

Approval checkpoints

High-impact actions stop and wait for you

Purchases, submissions, messages, deletions and settings changes are listed as proposed actions, each with the checkpoint the policy engine added and why.

  • Risk classified by rules you can read, not by the model
  • Unbounded scope is blocked before export
  • Sensitive fields stay as named placeholders until run time
Proposed agent actions with the approval checkpoints the policy engine requires

Spec risk auditPro

Check every spec against the OWASP LLM Top 10 risks agents hit most

Prompt injection, sensitive information disclosure, improper output handling, excessive agency and unbounded consumption: each scored pass, warn or fail before an agent ever runs the spec.

  • Findings ranked by severity, with the evidence from the spec
  • A concrete fix for every finding
  • Deterministic checks: the same spec always gets the same audit
OWASP Top 10 for LLM Applications risk audit of a generated specification

Agent context exportPro

One spec, every coding agent

Export the same spec as AGENTS.md, CLAUDE.md or a Cursor rule, so Codex, Claude Code, Cursor and browser agents all read the same instructions.

  • Keeps the trust boundary wording in every format
  • Copy or download each file
  • Compare versions side by side
A specification exported as AGENTS.md, with CLAUDE.md and Cursor rule tabs

Readable output

A spec your team can review like code

Objective, scope, allowed domains, steps with verification, recovery paths and the final report format, rendered as clean Markdown with a stable version hash.

  • Markdown, JSON and platform-ready formats
  • Send any spec to the Markdown workbench
  • Save requests as templates and reuse them
The generated task specification rendered as Markdown

Built for the people who answer for what the agent did

Engineers write the spec, product owners review the checkpoints, and operations decide what runs. Everyone reads the same document.

A software engineer reviews a task checklist on a large monitor in a loft office at dusk
Engineers turn a ticket into a spec with allowed domains and checks before handing it to an agent.
A product manager and a senior engineer review approval checkpoints together on a laptop
Product owners review the approval checkpoints instead of reading prompts.
An operations lead holds a tablet and decides whether to approve an automated task
Operations leads approve only the actions the spec says need a human.

Photos are AI-generated illustrations, not customers.

Review a spec from your phone

Check the assurance score and the checkpoints when an approval request reaches you away from your desk.

SpecStudio on a phone, showing the spec assurance card

a DevSpeak Labs product

One plan, four apps.

SpecStudio is one of four DevSpeak Utilities apps for the inputs to AI agents: specs, system prompts, technical text and prompt tests. Each app is free to use, and one Utilities Pro plan ($5.99/month or $59.99/year) unlocks Pro in all four.

  • SpecStudio's Studio: a travel-booking request with allowed domains, constraints and success criteria, ready to compile into a spec

    SpecStudio

    Agent task specs

    You are here

    Turn a rough request into a safety-gated spec a browser agent can follow.

    Pro adds in SpecStudio

    • Spec risk audit (OWASP LLM Top 10)
    • Export to AGENTS.md, CLAUDE.md and Cursor rules
    • Live page grounding
    • Pro template pack
  • Prompt Guardian's refinement dashboard: a support-agent system prompt in the editor with a clean static scan

    Prompt Guardian

    Gated prompt governance

    Attack, score and gate every system prompt change before it ships.

    Pro adds in Prompt Guardian

    • Advanced attack packs
    • OWASP LLM Top 10 (2025) report
    • Regression guard for CI
    Open Prompt Guardian
  • Dev Text Forge's workbench: an informal note about a deploy script in the source pane, next to the optimized output pane and quality gates

    Dev Text Forge

    Dev text optimizer

    Rewrite developer text into clear technical English that keeps every token.

    Pro adds in Dev Text Forge

    • Agent context mode
    • Token diet report
    • Batch optimize
    Open Dev Text Forge
  • PromptForge's tool directory: prompt-engineering tools such as System Prompt Maker, Prompt Optimizer and Image to Prompt

    PromptForge

    Prompt engineering toolbox

    A prompt-engineering toolbox: generators, a linter, tests and a workspace.

    Pro adds in PromptForge

    • Prompt linter
    • One-click prompt hardening
    • Assertion-based prompt tests
    Open PromptForge

Free to start. One price for Pro everywhere.

Use SpecStudio free for as long as you like. When you want the Pro features, one plan covers all four utilities.

Prices in USD, taxes included. One plan for all four apps.

Free

Everything SpecStudio does today. No card, no trial clock.

$0forever

Sign in with Google or email.

  • Spec generation with the deterministic policy pass
  • Spec assurance score and validation findings
  • Approval checkpoints and missing-information questions
  • Markdown, JSON and platform exports
  • Templates, saved requests and the Markdown workbench
Start free
Recommended

Utilities Pro Pro

Pro features in SpecStudio and in the other three utilities.

$5.00per month

Billed $59.99 once a year. Save 17% versus monthly.

7-day Pro trial, no card required

Sign in with Google and it starts by itself: every Pro feature in all four apps, with up to 30 Pro actions a day (100 in total) and 3 premium-model runs a day (5 in total). Then you're back on Free automatically; nothing to cancel.

One trial per email. Not available to emails that already had Utilities Pro, and new trials are limited each day.

  • Everything in Free
  • Spec risk audit (OWASP LLM Top 10)
  • Export to AGENTS.md, CLAUDE.md and Cursor rules
  • Live page grounding
  • Premium models
  • Pro template pack
  • Pro editor: AI assist, cloud history, HTML/PDF export, token budget

Pro in all four apps

  • SpecStudioThis app
  • Prompt GuardianGated prompt governance
  • Dev Text ForgeDev text optimizer
  • PromptForgePrompt engineering toolbox
Free and Utilities Pro compared, app by app
FeatureFreePro
In SpecStudio
Spec generation with the deterministic policy pass
Spec assurance score and validation findings
Approval checkpoints and missing-information questions
Markdown, JSON and platform exports
Templates, saved requests and the Markdown workbench
Spec risk audit (OWASP LLM Top 10)Score every spec 0-100 against the OWASP Top 10 for LLM Applications 2025 (prompt injection, sensitive data, output handling, excessive agency, unbounded consumption), with the evidence and a concrete fix for each finding.
Export to AGENTS.md, CLAUDE.md and Cursor rulesRender a validated spec as AGENTS.md, CLAUDE.md and a Cursor rule (.cursor/rules/*.mdc), ready to copy or download, regenerated from the spec so they never go stale.
Live page groundingGround a spec in the live content of the page the agent will operate on, read server-side and passed to the model as untrusted reference data, never as instructions.
Premium modelsThe large models in every model picker: Claude Opus 5.5, Claude Sonnet 5.5, Mistral Large 4 and Mistral Medium 3.5. Claude Haiku 5.5 (the default), Mistral Medium 3.1, Devstral 2, DeepSeek and the other fast models stay free for everyone.
Pro template packSeven ready-made specs: checkout QA, listing extraction, multi-step forms with approval checkpoints, accessibility and account-settings audits, price monitoring and inbox triage.
Pro editor (every app)
AI writing assist in the editorRewrite, tighten, fix, expand or turn any selection into a checklist.
Cloud version historyEvery save kept in the cloud, with a side-by-side diff and one-click restore.
HTML and PDF exportShare polished documents with stakeholders who never open a Markdown file.
Token and context budgetSee how much of each model's context window a document uses before you paste it.
In Prompt Guardian
Prompt Guardian: everything free todayAttack, score and gate every system prompt change before it ships.
Advanced attack packs
OWASP LLM Top 10 (2025) report
Regression guard for CI
In Dev Text Forge
Dev Text Forge: everything free todayRewrite developer text into clear technical English that keeps every token.
Agent context mode
Token diet report
Batch optimize
In PromptForge
PromptForge: everything free todayA prompt-engineering toolbox: generators, a linter, tests and a workspace.
Prompt linter
One-click prompt hardening
Assertion-based prompt tests
  • Free stays free. Pro only adds; nothing you use today moves behind the paywall.
  • Cancel anytime from Settings → Billing. Pro stays active until the end of the period you paid for.
  • One subscription, four apps. In another app, sign in with Google using the same email and Pro is already on.
  • Secure checkout by Dodo Payments, merchant of record. It handles sales tax and VAT; prices include them.

Questions, answered

Does SpecStudio run the browser task?

No. It writes the specification. You run it on the agent platform you already use, and the spec tells that agent where to stop and ask.

Is the free plan really free?

Yes. Spec generation, the policy pass, exports, templates and the workbench are free with no card and no trial period. Utilities Pro only adds features.

What does Utilities Pro add here?

The OWASP LLM Top 10 risk audit, AGENTS.md, CLAUDE.md and Cursor rule exports, live page grounding, the Pro template pack, the Pro editor tools, and the premium models: Claude Opus 5.5, Claude Sonnet 5.5, Mistral Large 4 and Mistral Medium 3.5. The same plan also unlocks Pro in the other three utilities.

Do you store my credentials or personal data in specs?

No. Sensitive fields are written as named placeholders such as {{TRAVELER_FULL_NAME}} that you fill in at run time. The privacy policy lists exactly what is stored.

Which models write the spec?

Claude Haiku 5.5 by default, with Mistral Medium 3.1 via OpenRouter as the fallback. You can also pick a model: Claude Haiku 5.5, Mistral Medium 3.1 and Devstral 2 are free, and so are DeepSeek, Xiaomi MiMo and Zhipu GLM, which run only when you pick them and are processed in China. Claude Sonnet 5.5, Claude Opus 5.5, Mistral Large 4 and Mistral Medium 3.5 come with Utilities Pro. Whichever model you pick, your request is sent as data inside tags, never concatenated into instructions.

Can I cancel Utilities Pro?

Anytime, from Settings → Billing. You keep Pro until the end of the period you paid for, then the free plan continues.

Give your next agent task a spec it cannot misread.

Start free. Upgrade when you want the risk audit and agent-context exports, in this app and the other three.

Compile a spec free