# agentmark

> Markdown for the AI era

- **URL**: https://www.freshcrate.ai/projects/agentmark
- **Author**: agentmark-ai
- **Category**: Prompt Engineering
- **Latest version**: `@agentmark-ai/ui-components@0.6.4` (2026-06-05)
- **License**: MIT
- **Source**: https://github.com/agentmark-ai/agentmark
- **Homepage**: https://docs.agentmark.co/agentmark/
- **Language**: TypeScript
- **GitHub**: 345 stars, 21 forks
- **Registry**: github
- **Tags**: `agents`, `llms`, `observability`, `opentelemetry`, `prompt`, `prompt-engineering`, `prompt-management`, `typescript`

## Description

Markdown for the AI era

## Recent releases

| Version | Date | Urgency | Changes |
| --- | --- | --- | --- |
| `@agentmark-ai/ui-components@0.6.4` | 2026-06-05 | High | ### 🧱 Updated Dependencies  - Updated @agentmark-ai/prompt-core to 0.7.0 |
| `@agentmark-ai/ui-components@0.6.2` | 2026-05-22 | High | ### 🧱 Updated Dependencies  - Updated @agentmark-ai/api-schemas to 0.3.0 - Updated @agentmark-ai/prompt-core to 0.5.0 |
| `create-agentmark@0.10.7` | 2026-05-14 | High | ### 🩹 Fixes  - **Python `FileLoader` now matches the TypeScript `FileLoader` API.** ([#596](https://github.com/agentmark-ai/agentmark/pull/596))    - `agentmark-prompt-core`: `FileLoader` constructor now takes a single `build_dir` positional argument — pointing directly at the build output directory (`agentmark build`'s output), matching `new FileLoader('./dist/agentmark')` in TS. The argument defaults to `"./dist/agentmark"`, so `FileLoader()` from a project root with the conventional build l |
| `@agentmark-ai/ui-components@0.6.1` | 2026-05-12 | High | ### 🧱 Updated Dependencies  - Updated @agentmark-ai/api-schemas to 0.2.0 |
| `agentmark-sdk@0.2.1` | 2026-04-16 | High | ### 🩹 Fixes  - fix: set explicit User-Agent on OTLP span exports to bypass Cloudflare BIC ([#584](https://github.com/agentmark-ai/agentmark/pull/584))    Cloudflare's Browser Integrity Check rejects requests bearing the default   `Python-urllib/*` User-Agent with HTTP 403 (error code 1010). `JsonOtlpSpanExporter`   uses `urllib.request.urlopen` without setting a UA, so every trace export through   a Cloudflare-proxied zone (api.agentmark.co, api-stg.agentmark.co) was silently   rejected before |
| `@agentmark-ai/ui-components@0.5.2` | 2026-04-15 | High | ### 🩹 Fixes  - Unify scorer storage format across the eval runner and annotation UI, rename the client `scores` option back to `evals`, and refresh the model catalogue. ([#581](https://github.com/agentmark-ai/agentmark/pull/581))    ### @agentmark-ai/prompt-core    - `ScoreSchema.categorical.categories` is now `Array<{ label: string; value: number }>` instead of `string[]`. Each category carries its own numeric value used when posting scores. Consumers constructing categorical score configs mu |
| `@agentmark-ai/sdk@1.1.1` | 2026-04-15 | High | ### 🧱 Updated Dependencies  - Updated @agentmark-ai/prompt-core to 0.4.1 |
| `@agentmark-ai/prompt-core@0.4.1` | 2026-04-15 | High | ### 🩹 Fixes  - Unify scorer storage format across the eval runner and annotation UI, rename the client `scores` option back to `evals`, and refresh the model catalogue. ([#581](https://github.com/agentmark-ai/agentmark/pull/581))    ### @agentmark-ai/prompt-core    - `ScoreSchema.categorical.categories` is now `Array<{ label: string; value: number }>` instead of `string[]`. Each category carries its own numeric value used when posting scores. Consumers constructing categorical score configs mu |
| `@agentmark-ai/model-registry@0.2.3` | 2026-04-15 | High | ### 🩹 Fixes  - Unify scorer storage format across the eval runner and annotation UI, rename the client `scores` option back to `evals`, and refresh the model catalogue. ([#581](https://github.com/agentmark-ai/agentmark/pull/581))    ### @agentmark-ai/prompt-core    - `ScoreSchema.categorical.categories` is now `Array<{ label: string; value: number }>` instead of `string[]`. Each category carries its own numeric value used when posting scores. Consumers constructing categorical score configs mu |
| `@agentmark-ai/mastra-v0-adapter@1.3.2` | 2026-04-15 | High | ### 🩹 Fixes  - Unify scorer storage format across the eval runner and annotation UI, rename the client `scores` option back to `evals`, and refresh the model catalogue. ([#581](https://github.com/agentmark-ai/agentmark/pull/581))    ### @agentmark-ai/prompt-core    - `ScoreSchema.categorical.categories` is now `Array<{ label: string; value: number }>` instead of `string[]`. Each category carries its own numeric value used when posting scores. Consumers constructing categorical score configs mu |

## Dependency audit

- **Score**: 100/100
- **Total deps**: 12
- **Resolved**: 12
- **Unresolved**: 0
- **License conflicts**: 0
- **Warnings**: 0
- **Scanned**: 2026-05-18

## Citation

- HTML: https://www.freshcrate.ai/projects/agentmark
- Markdown: https://www.freshcrate.ai/projects/agentmark.md
- Dependencies JSON: https://www.freshcrate.ai/api/projects/agentmark/deps

_Generated by freshcrate.ai. Indexes github releases for AI-agent ecosystem packages._