Use promptev’s AI with no keys at all, or bring your own. We manage infrastructure, memory, sessions, context ingestion, auto-indexing, embedding, retrieval, tools, and governance: no RAG overhead, no pipeline to build. Every feature is included from the first day; you buy credits when you need them, and nothing else.
1 credit = 1 AI action. Most queries use 2–5 credits on your own model key, about 15 on PromptevAI. Ingestion: 1 credit/page for documents & PDFs, 1 credit/slide for presentations, 1 credit/MB for spreadsheets & data files. See full breakdown ↓ · Estimate your cost ↓
Last updated: October 23, 2026
Start building now.
No credit card required · 1,000 credits free for 30 days · No subscription
Transparent, predictable pricing. You know exactly what you're paying for.
Every request uses promptev’s orchestration, logging, and streaming infrastructure, on your own model key or on PromptevAI.
Semantic search across your Context Packs — semantic search. Base + context = 2 credits.
Graph-based retrieval with entity linking and relationship traversal. Base + graph = 6 credits.
An agent calling Jira, Slack, Gmail or SharePoint: 1 credit per tool, additive. The same for each call an app you connected (Claude, ChatGPT, Cursor) makes to build or use your agents.
When an agent runs code over a spreadsheet (sums, pivots, charts), on top of the tool call. Nothing if it fails on our side.
When an agent answers with a PromptevAI model, the model’s cost is deducted from your credits. The credits above stay the same. You see the model credits on every answer and in Usage.
Base (1) + context retrieval (1) + model (≈13) ≈ 15 credits, about 7.5¢. On the default model. A short answer costs less, a long document more, and another PromptevAI model can cost more or less.
Connect OpenAI, Anthropic, Google, DeepSeek or another provider and answers use your key. Your provider bills you for the model; promptev charges only the credits above.
A 10-page PDF costs 10 credits. Includes text extraction, structured data processing, and OCR for scanned pages. Charged once per sync.
Each image (JPEG, PNG, TIFF, etc.) counts as 1 page. Includes text extraction and structured data processing.
Word documents charged per page, just like PDFs. A 12-page report = 12 credits. Includes text extraction, image processing, and structured data.
PowerPoint files charged per slide. A 17-slide deck = 17 credits. Includes text extraction, chart/image processing, and structured data.
Excel, CSV, code files, text — charged by file size (minimum 1 credit). A 3 MB spreadsheet = 3 credits. A 200 KB CSV = 1 credit.
A voice note sent to an agent (for example on WhatsApp) is transcribed so the agent can read it.
Entity extraction, relationship mapping, and topic clustering. Only applies to graph-mode packs — semantic packs skip this.
Base (1) + context retrieval (1) + Jira (1) + Slack (1) + email (1) = 5 total.
PDF (20) + PPTX slides (17) + DOCX pages (12) + images (2) = 51 total. All processing included. Embedding always free.
Playground testing counts towards credits. Every request (chat, playground, API, scheduled, webhook) uses credits. Embedding is always free.
Each piece is separate. Until you configure your own for a piece, promptev’s providers do that work for you. Ingestion and retrieval credits are the same either way.
Ours: PromptevAI, its cost deducted from your credits. Yours: your own key, set under AI providers. Your own model covers answers only, not document reading or search.
Ours: included in the ingestion credits. Yours: your own vision or OCR model, set in the context engine settings.
Ours: included in the ingestion credits. Yours: your own extraction model, set in the context engine settings.
Ours: always free. Yours: your own embedding model, set in the context engine settings.
Ours: promptev’s database. Yours: your own PostgreSQL with pgvector, set in the context engine settings.
The providers promptev uses for the pieces you do not bring are listed on our Sub-Processors page.
Put in how you expect to use promptev. The credits and the price use the rates above.
Everything is included. Answers run on PromptevAI (≈13 credits a typical answer on the default model, deducted from your credits). Document reading (OCR), extraction, embeddings run on PromptevAI, and the database is promptev’s. Nothing to set up and no other bill.
Your model costs are not included. Answers use your own model key, so your provider (OpenAI, Anthropic, Google and others) bills you for them. We never mark it up. Document reading, extraction, embeddings and the database are promptev’s.
Your provider costs are not included. Answers, document reading (OCR), extraction and embeddings run on your own providers, who bill you directly, and your knowledge is stored in your own PostgreSQL. promptev charges only the credits shown. Each piece is set up separately: your model under AI providers, the rest in the context engine settings.
First month:
The Context Graph is charged 1 credit per 8 chunks and 1 per topic cluster. Neither is known before a file is read, so its line is an estimate, set high on purpose. Embedding is always free.
Start Free — 1,000 Credits →What companies typically spend to get what Promptev gives you out of the box.
Building and maintaining RAG pipelines, integrations, prompt management, and context retrieval — one person, full-time. Promptev starts free.
Pinecone, Weaviate, or pgvector hosting + embedding pipelines + monitoring. Included in Promptev, for everyone.
Building context pipelines, tool orchestration, auth, and governance from scratch — or just connecting your data and going live.
Every AI request = 1 credit. Context retrieval adds +1 credit (or +5 for Context Graph). Each tool call = +1 credit. File ingestion: 1 credit per page for PDFs & documents, 1 credit per slide for presentations, 1 credit for images, 1 credit per MB for spreadsheets & data files. Example: query + context + 3 tools = 5 credits. Embedding is always free.
No. There are no plans to hold anything back behind. Roles and permissions, the context graph, every integration, single sign-on, approvals, the audit trail, unlimited projects and agents — all of it is in every workspace, including one that has never paid. The only thing you buy is credits. Enterprise buys a custom deployment, onboarding and support — a place to run and people to call, not features.
Everything, including the AI model if you want: answers, document reading (OCR), extraction and embeddings run on PromptevAI with no key. We handle infrastructure, memory, sessions, context ingestion, auto-indexing, embedding, document processing, retrieval, tool orchestration, and governance, with no RAG pipelines to build or maintain. Embedding is always free. Bring your own model, OCR, extraction, embeddings or database whenever you want; each is separate.
We e-mail you at 80%, 90% and 100% of your last top-up, and auto top-up can refill for you before you run low. When the balance is empty, a new question or action waits until you top up; your agents, conversations and data are all kept.
Yes. Every request uses credits — playground, chatbot, API, scheduled, or webhook. No separate test mode. Only embedding is free.
Bring Your Own Keys. Connect your model provider: OpenAI, Anthropic (Claude), Google Gemini and DeepSeek built in, 34 more ready-made (Mistral, Groq, OpenRouter, xAI, Azure OpenAI, Google Vertex AI, Amazon Bedrock and others), or any compatible endpoint of your own. Encrypted with AES-256. Your own model key covers answers only. To use your own OCR, extraction, embeddings or database too, set each in the context engine settings. Without a key, answers run on PromptevAI and are charged in credits.
Yes. Special pricing for early-stage startups and nonprofits. Contact us.
Our team can help scope your use case and recommend the right tier.
No credit card required · 1,000 credits free for 30 days · No subscription