Middle Mann  /  Compare  /  AI Workspaces  /  Claude

Cherry is the self-hosted alternative to Claude.

Claude ships one vendor's models (Opus, Sonnet, Haiku), a flat Projects UI and rolling weekly usage caps that Anthropic doesn't fully surface, with every conversation and artifact stored on Anthropic's infrastructure. Cherry ships the same Claude models via the Anthropic Messages API, wrapped in an infinite nested folder tree, the Pins/Clipboard/Snippets three-tier reuse system, per-message cost attribution and private MinIO file storage. OpenAI's GPT models are one click away in the same picker. Self-hosted end to end. Deploy on your hardware, run it as your own SaaS, do whatever you want. You own the whole stack.

Flat
Claude Projects sit flat. Cherry ships infinite nested folders.
Silent
Claude Pro and Max hide rolling weekly caps. Cherry logs every call to the API.
2
Two provider adapters ship today. GPT sits alongside Claude in one workspace.
Compare
Why It Exists

Four places Claude's consumer UI runs out of room.

Claude has the best reasoning in the field for a lot of tasks. It also has a consumer chat UI that pretends professionals only run one project at a time and never need to see how much any of it costs.

01

Claude Projects are flat containers

Claude Projects group related conversations under a per-project system prompt and shared files. Pro plans get unlimited Projects, but Projects themselves don't nest. There are no sub-projects and no way to organize a research thread across nested subtopics. Cherry's Folder entity is self-referential, drag-moveable, indexed for global search, with unbounded nesting depth. That's real filesystem structure for people running dozens of parallel workstreams.

02

Silent rolling caps in the hosted UI

Claude Pro throttles on rolling weekly limits. Max 5x and Max 20x multiply those limits but keep the same opaque window. The UI shows "you're approaching your limit" without a real budget you can read. Cherry hits the Anthropic Messages API directly and logs every call as a UsageEvent with tokens and dollar cost. The Recharts dashboard shows exactly where you are against any budget you set.

03

Artifacts are trapped in the chat

Claude's Artifacts render beautifully inside a single conversation and vanish the moment you close the tab. Cherry's Pins bookmark any message in place. Clipboard saves passages per chat and lets you drag-reorder them. Snippets promote those passages into project-scoped reusable text with tags. Every good response becomes a reusable asset instead of a lost tab.

04

Files and history live on Anthropic's servers

Every upload sits in Anthropic's storage layer. Every conversation lives in their database. Cherry ships private MinIO with local-disk fallback, presigned URLs with short TTL, filename sanitization and zero bucket-path exposure. Postgres holds the conversation history on your infrastructure. Nothing leaves your stack.

What Cherry Ships

Anthropic's models. Workspace bones Anthropic didn't ship.

Cherry hits the Anthropic Messages API directly, so every Opus, Sonnet and Haiku conversation still runs against the model you're used to. The rest is different. A nested folder tree that keeps going where Claude Projects stop. Rolling weekly caps that become an actual budget you can read. Artifacts that persist as Pins, Clipboard entries and Snippets instead of dying with the tab. Files that live on your MinIO, not on Anthropic's servers. And a Compare modal that fires GPT next to Claude on the same prompt.

Past Claude Projects
01Folders inside projects. Claude Projects don't nest. Cherry's Folder entity is self-referential with unbounded depth
02Artifacts that persist. Claude Artifacts die with the tab. Cherry Pins and Clipboard save the response as a first-class asset
03Snippets across projects. Claude's per-Project prompt stays there. Cherry Snippets tag, scope and inject to any chat anywhere
04Bulk file ingest. Claude uploads one file per turn. Cherry replicates an entire OS folder tree transactionally in one action
Beyond Anthropic's Silo
01Route past Claude. Claude only serves Claude. Cherry ships Anthropic + OpenAI today with a typed adapter for Google, xAI, more
02Fire at every model. Claude has no side-by-side compare. Cherry's Compare modal fires one prompt at every provider at once
03Presets across providers. Claude's Project prompt is Claude-scoped. Cherry Presets bundle model + temp + prompt across every adapter
04Standard streaming. Claude uses its own transport. Cherry streams via standard Server-Sent Events with real cancel-mid-stream
What Anthropic Won't Let You See
01Cost per prompt. Claude Pro/Max hides token spend behind rolling caps. Cherry logs tokens + dollars + latency on every message
02File retention. Claude files sit on Anthropic. Cherry stores in MinIO on your infrastructure with presigned URLs, short TTL
03Where the cap actually is. Pro / Max 5x / Max 20x hide the real rolling limit. Cherry hits the API directly with transparent quotas set per plan
Side by Side

Claude vs Cherry.

Same Claude models via the same Anthropic Messages API. Everything above the model layer is different.

Feature Claude Cherry
Model providersAnthropic onlyAnthropic + OpenAI shipping, pluggable adapter for any provider
OrganizationProjects (flat, don't nest)Infinite nested folder tree, self-referential parents
Per-message bookmarksNoPins with scroll-to-highlight animation
Per-chat clipboardNoDrag-reorderable saved passages, promote to Snippet
Cross-chat snippetsNo first-class equivalentProject-scoped, tagged, one-click injection to any chat
Artifacts persistenceIn-chat only, tab-scopedPinnable, promotable to Snippet, searchable via global search
Model presetsPer-Project system promptAgent Library. Presets with model + temp + system prompt + tier
Model comparisonNot supportedFire one prompt at multiple models side by side
Cost per promptNot shownTokens in, tokens out, dollar cost, latency on every message
Cost attributionPer seat monthly aggregatePer user, per workspace, per chat, per model
Usage capsSilent rolling window in Pro / MaxAPI-direct. Transparent quotas set per plan
Usage dashboardNone for end usersRecharts. Today / 7d / 30d / billing cycle
Bulk folder uploadNoReplicates an entire OS folder tree transactionally
File storageAnthropic infrastructureSelf-hosted MinIO with local-disk fallback, presigned URLs
Conversation historyAnthropic-ownedYour Postgres. Exportable, portable
Streaming transportProprietaryServer-Sent Events with cancel-mid-stream
Team workspacesClaude Team (Standard $20-25/seat, Premium $100-125/seat, 2-150 users)Multi-workspace with Owner / Admin / Member roles, no per-seat cap
SSOTeam tier and upAvailable in Enterprise, configurable on your infrastructure
Keyboard shortcutsLimited18 mapped across Navigation / Chat / Editor / Files
ConnectorsGrowing catalogGitHub + Google Drive shipping, Notion / Slack / Jira on roadmap
AppsArtifacts, Claude Code, Claude Cowork (all in-Claude)Canvas, Code Interpreter (beta), Image Generation, modular
Outbound APINot a first-class product surfaceCherry outbound API for third-party integration
DeploymentSaaS onlySelf-hosted. On-prem, cloud, run as your own SaaS. You decide
OwnershipVendor-ownedYou own the code, the data, the deployment
FAQ

Frequently asked questions.

Can we still use Claude models in Cherry?

Yes. Anthropic is a first-class shipping provider via the Anthropic Messages API. Every prompt that used to go through Claude's UI hits the same models. The difference is that every message is logged as a UsageEvent with tokens, dollar cost and latency attributed to the specific user, workspace, chat and model. And GPT is one click away in the same picker with the same Preset semantics.

What about Claude Projects and per-project instructions?

Cherry replaces Projects with the Folder tree plus Agent Library Presets. A Preset captures the model, temperature and system prompt. Folders group the chats. Snippets hold project-scoped reusable text with tags. The whole configuration is portable and exportable and works across every provider adapter, not only Claude.

What happens to Artifacts?

Cherry has a Canvas app that gives you a live rendering surface for generated code, plus a Code Interpreter beta and an Image Generation app. On top of that, any message can be Pinned in place, saved to the Clipboard for later drag-reorder or promoted to a Snippet for project-wide reuse. The good responses stop being tab-scoped and start being assets.

Where do uploaded files live?

Files live in MinIO on your infrastructure, with a local-disk fallback if MinIO isn't wired. The storageKey is stored in the File record. The raw bucket path is never exposed. Every file access goes through a short-TTL presigned URL generated per request. Filenames are sanitized. If you shut Cherry down tomorrow, the files stay in your bucket. There's no vendor deprovisioning risk.

How does Cherry stream responses?

Cherry streams via Server-Sent Events over HTTP. Every provider adapter implements the streaming interface, so tokens appear the moment the model emits them. The Stop button cancels mid-stream. SSE is a standard browser API. There's no proprietary transport and no vendor lock-in.

How is Cherry deployed?

Cherry is self-hosted. Docker Compose orchestrates nginx, the Next.js frontend, the Express backend, Postgres, Redis and MinIO. Deploy it on your own hardware for internal use, run it in your cloud or spin it up as your own SaaS product. You own the code, the data and the deployment.

What does Cherry cost?

Contact sales for pricing. Every engagement is scoped to how you plan to deploy, who your users are, which provider adapters you need lit up and what support terms fit your organization.

Start a Conversation

Ready to move off Claude and own the workspace?

Tell us how your team uses Claude today, how many concurrent projects you juggle and where it hits the wall. We'll respond within one business day with an honest read on whether Cherry fits.

Build yours

Want to own your AI workspace?

[email protected]