Prompt Assay
Prompt Assay is a workbench for authoring, refining, versioning, testing, and shipping LLM prompts and Agent Skills.
What Prompt Assay is
Prompt Assay is a workbench for authoring, refining, versioning, testing, and shipping LLM prompts and Agent Skills. Its critique tool rates prompts on clarity, completeness, structure, technique use, robustness, and efficiency, with written explanations. An in-editor AI pair can brainstorm, critique, improve, rewrite, compare, and convert content. Users can connect their own Anthropic, OpenAI, or Google models, run one prompt across two to five models in parallel, and compare responses, latency, token use, and cost; optional LLM-as-Judge scoring is available. Agent Skills can be built as multi-file bundles, critiqued, and evaluated across Claude, GPT, and Gemini. Imports accept Markdown, Anthropic JSON, OpenAI JSON, or raw text. The Prompt Assay bundle keeps versions, tags, fragments, and metadata. The free plan has one seat and 250 calls monthly, with seven initial platform-funded calls before BYOK is needed for ongoing use. Solo costs $49 a month and includes API and SDK access; the public REST API and TypeScript SDK require Solo or higher.
Good
- Scores prompts across six criteria with written reasons.
- Compares outputs from two to five models.
- Supports multi-file Agent Skill bundles and behavioral evaluations.
- Imports Markdown, Anthropic JSON, OpenAI JSON, and raw text.
- Free plan includes 250 calls per month.
Watch out
- Free tier is limited to one seat.
- Ongoing free-tier use requires BYOK.
- REST API and TypeScript SDK require Solo or higher.
- Solo costs $49 a month.
Who it suits
Prompt Assay suits prompt engineers and teams sharing prompt libraries, as well as researchers comparing models and developers retrieving prompts through an API or SDK. It is useful for users who want to manage prompts and Agent Skills through versioning, testing, and evaluation.
Verdict
Prompt Assay brings prompt and Agent Skill authoring, critique, testing, and versioning into one workbench. The free tier has limits, while API and SDK access requires a paid Solo tier or above.
Plans and prices
From Prompt Assay's own pricing page, checked 7 Oct 2001.
| Plan | Price | What it gives you |
|---|---|---|
| Solo | $49 a month | Unlimited AI calls on your keys · full prompt and Skill version history · public REST API and TypeScript SDK · single seat · $49/ month |
| Team | $99 a month | 3 to 15 seats · shared org library and evaluation suites · owner, admin, and member roles · $99/ seat / month; minimum 3 seats; annual billing available |
| Free | — | 7 platform-funded calls to start · 250 AI calls a month · single seat · 7-day version view · BYOK for ongoing use · $0/ month |
| Enterprise | Priced on request | Unlimited seats · SAML SSO · custom tier controls · data processing agreement · priority support · Custom contract |
The facts
Each one read from Prompt Assay's own pages, with the date it was checked.
What it does
- Agent Skills
- Users can author multi-file Agent Skill bundles, critique them, and run Behavioral Evals across Claude, GPT, and Gemini.checked 7 Oct 2001
- AI actions
- The in-editor AI pair offers brainstorm, critique, improve, rewrite, compare, and convert actions.checked 7 Oct 2001
- API
- The public REST API and TypeScript SDK require Solo tier or higher; API requests are authenticated with an API key.checked 7 Oct 2001
- Critique
- Its prompt critique scores clarity, completeness, structure, technique usage, robustness, and efficiency, with written reasons.checked 7 Oct 2001
- Data retention
- Provider responses are generally not retained server-side, with evaluation test outputs and Skills Behavioral Eval results as stated exceptions.checked 7 Oct 2001
- Imports and exports
- Prompt imports support Markdown, Anthropic JSON, OpenAI JSON, and raw text; the Prompt Assay bundle preserves versions, tags, fragments, and metadata.checked 7 Oct 2001
- Intended users
- The maker describes the product for prompt engineers, teams sharing prompt libraries, researchers evaluating models, and developers retrieving prompts through an API or SDK.checked 7 Oct 2001
- Key handling
- Provider API keys are encrypted at rest, never appear in logs, and are used only for LLM calls the user triggers.checked 7 Oct 2001
- Model comparison
- Users can run one prompt across 2–5 BYOK models in parallel and compare outputs, latency, tokens, and cost, with optional LLM-as-Judge scoring.checked 7 Oct 2001
- Model providers
- The workbench supports BYOK connections to Anthropic, OpenAI, and Google models.checked 7 Oct 2001
- Product
- Prompt Assay is a workbench for authoring, refining, versioning, testing, and shipping LLM prompts and Agent Skills.checked 7 Oct 2001
- Training and certifications
- Prompt Assay says it does not train on customer content and does not currently hold SOC 2 or ISO 27001 attestations.checked 7 Oct 2001
Money
- Free tier limit
- Free includes 250 calls per month and one seat; seven initial platform-funded calls are available before BYOK is needed for ongoing use.checked 7 Oct 2001
The company
- Maker
- Jon Lasley founded Prompt Assay; the About page describes his enterprise technology and applied AI experience but does not state a founding year or headquarters.checked 7 Oct 2001
Alternatives to Prompt Assay
- 1
Vellum
Free planOpen sourceAndroidiOSWebAPI8.7From $30 a month - 2
Langfuse
Free planOpen sourceWebAPILinuxSelf-hosted8.1From $29 a month - 3
PromptHelm
Free planAndroidiOSWebAPI8.0From $39 a month - 4
Agenta
Free planOpen sourceWebAPISelf-hosted7.9From $29 a month - 5
PromptFluent
Free planWebExtensionLinuxmacOS7.7From $14 a month - 6
Braintrust
Free planWebAPISelf-hosted7.6From $249 a month