- Home
- AI Red Teaming Tools
- PromptRedTeam
PromptRedTeam
What it is
PromptRedTeam scans user inputs for attacks before they reach a large language model. Its checks cover direct injection, role manipulation, zero-width injection, delimiter injection, and encoded payloads. The project describes direct-injection checks that look for override terms, conflicting instructions, and prompt structure, while role checks recognize jailbreak patterns, role-play scenarios, and bypass attempts. For hidden or encoded text, it detects zero-width characters, normalizes Unicode, analyzes Base64 and other encodings, and can decode suspicious content before rescanning it. Delimiter checks look for context-breaking markers, Markdown misuse, and structural manipulation. Applications can use the hosted API, for which the homepage demonstrates a POST request to https://api.promptredteam.com/test, or self-host the project. The live tester accepts up to 1,000 characters and allows 10 requests per minute; self-hosting provides unlimited usage. The site offers Python, JavaScript, and cURL implementation examples. Its suggested workflow is to scan inputs, then block threats or log them for review. PromptRedTeam is open source under the MIT License and is available as web, API, and self-hosted deployment. The project is free: the hosted demo is limited to 10 requests per minute, while self-hosting allows unlimited use.
For and against
For
- Checks five listed attack categories
- Can decode and rescan encoded text
- Self-hosting provides unlimited usage
- Offers Python, JavaScript, and cURL examples
- Open source under the MIT License
Against
- Hosted demo allows 10 requests per minute
- Live tester input capped at 1,000 characters
The verdict
PromptRedTeam offers hosted and self-hosted ways to scan inputs before they reach an LLM. The live tester has request and character limits; self-hosting is listed for unlimited usage.
Who it suits
PromptRedTeam suits developers integrating input checks into applications that send user text to LLMs. It also suits teams that want a self-hosted option or examples in Python, JavaScript, or cURL.
Plans and prices
Open source
Hosted demo: 10 requests/min · Self-hosted: unlimited usage
Read 9 Oct 2026 · promptredteam.com
On the record
- Attack coverage
- It detects direct injection, role manipulation, zero-width injection, delimiter injection, and encoded payloads.promptredteam.com · read 8 Oct 2026
- Company details
- The opened maker page does not state the maker's name, headquarters, or founding year.promptredteam.com · read 9 Oct 2026
- Delimiter detection
- The site says it detects context-breaking delimiters, markdown abuse, and structural manipulation.promptredteam.com · read 8 Oct 2026
- Detection categories
- It detects direct injection, role manipulation, zero-width injection, delimiter injection, and encoded payloads.promptredteam.com · read 9 Oct 2026
- Direct injection detection
- The site says direct injection detection uses pattern matching for override keywords, instruction conflicts, and prompt structure analysis.promptredteam.com · read 8 Oct 2026
- Hidden and encoded text
- The site describes zero-width character detection, Unicode normalization, Base64 and encoding analysis, and automatic decoding and rescanning.promptredteam.com · read 8 Oct 2026
- Hosted API
- The site demonstrates sending text to https://api.promptredteam.com/test using a POST request with JSON.promptredteam.com · read 9 Oct 2026
- Integration examples
- The site provides implementation examples for Python, JavaScript, and cURL.promptredteam.com · read 9 Oct 2026
- License
- The homepage describes PromptRedTeam as open source under the MIT License.promptredteam.com · read 8 Oct 2026
- Live testing limit
- The live demo allows 10 requests per minute and accepts up to 1,000 characters in its input field.promptredteam.com · read 9 Oct 2026
- Open source license
- The site describes the project as open source under the MIT License.promptredteam.com · read 9 Oct 2026
- Purpose
- PromptRedTeam scans user inputs for prompt injection before they reach an LLM.promptredteam.com · read 9 Oct 2026
- Rate limit
- The live tester allows 10 requests per minute and says deploying your own provides unlimited usage.promptredteam.com · read 8 Oct 2026
- Role manipulation detection
- The site says role manipulation detection recognizes known jailbreak patterns, role-play scenarios, and bypass attempts.promptredteam.com · read 8 Oct 2026
- Self-hosting
- The site says users can self-host for unlimited usage.promptredteam.com · read 9 Oct 2026
- Suggested use
- The homepage suggests scanning inputs before they reach an LLM, then blocking threats or logging them for review.promptredteam.com · read 8 Oct 2026
- Workflow
- The site says applications can scan inputs before they reach an LLM and block threats or log them for review.promptredteam.com · read 9 Oct 2026
