Every surface, every feature
One hybrid platform for local and cloud AI: workspace tools, agents, code and builds, data and memory, and a versioned studio to customise orchestration.
Flagship surfaces
- Maestro orchestration
A single chat that plans, retrieves, edits files, runs tools and schedules follow-ups. BYOK any model — local or cloud — and switch mid-conversation. - Multi Chat
Run the same prompt across two or more models side-by-side. Compare answers, costs, latency — pick the winner. - Agent Sessions
See every long-running agent at a glance: who is working, who is waiting, who is done. - Agent Mail
Agents send and receive structured messages from each other. Inbox for AIs. - Handoffs
When one agent finishes its part, it hands off to the next with full shared context. - Sources
Pick exactly which files, notes and folders Maestro sees in this conversation. - Pipeline roles
Assign a different model to each role — Router, Direct, Orchestrator, Fallback. Tune per workspace. - Tools
Every tool is opt-in. Flip on web_search, sql_query, edit_file… everything else stays off. - .md export
Every conversation has a clean markdown view — copy, share, version, diff. - Request / Response logs
Inspect the raw JSON for every model call. Debug prompts, count tokens, audit cost. - Right panel — Browser
Maestro can pop a live browser preview next to the chat to verify pages or scrape data. - Right panel — Code
Files Maestro is editing open side-by-side with the chat in a full Monaco editor. - Right panel — Execution
See terminal output from tool calls live, complete with green pass lines and stack traces. - BYOK keys
Bring your own API keys for any provider. Local models stay local, cloud models stay yours. - Model picker
Search and switch between local and cloud models in one keystroke — no provider menus. - Memory groups
Scope conversations to a project, customer, or playbook. Memories survive across sessions. - Sandbox modes
Run agent code on your laptop, a managed container, or a remote box. Same agent, different blast radius. - Guards
Block dangerous tool calls before they run — destructive shell, off-policy URLs, PII writes. Configurable per workspace. - Plan mode
For risky tasks, Maestro proposes a plan first and waits for explicit approval before touching anything. - Todo list
Long tasks break into checklists Maestro maintains as it works — you always know what is left. - File edit proposals
Every Maestro edit shows up as a reviewable diff. Accept, reject, or ask for a redo. - Agent plan view
Watch any agent decompose a goal into a tree of subtasks — done, running, pending — in real time. - Agent timeline
Every step an agent takes, with timestamps and durations. Full replay of any run. - Agent triggers
Wake an agent up on a webhook, a database event, a cron — or chain from another agent. - Workflows
Visual builder for chained agents and tools — drag, drop, connect, ship. - Model playground
Test any model in a sandbox prompt → response interface with full token / cost accounting. - Local servers
Run local inference on your machine or workstation. Routes are automatic — Maestro picks local when it can. - Cloud GPU
Rent A100 / H100 / L40 instances directly from the platform. Pay per hour, no contracts. - Training
Fine-tune or LoRA-adapt models on your own data — locally, on a workstation, or in the cloud. - Benchmarks
Side-by-side scores across MMLU, GSM8K, HumanEval and others. Pick a model with data. - Notes — Text
A modern rich text editor — Markdown, slash commands, code blocks, math, embeds. - Notes — Whiteboard
Sketch, diagram and connect ideas on an infinite canvas. Shapes, arrows, sticky notes. - Notes — Slides
Build presentations alongside your docs. Same workspace, same Maestro, no PowerPoint. - Notes — Diagrams
Mermaid diagrams live next to their source. Edit on one side, render on the other. - Notes — Research
Card-based view to collect sources, quotes and highlights as you research a topic. - Note access
Per-note visibility: private, workspace, or public. Granular per-member overrides. - Note versions
Every save is a version. Compare any two, roll back instantly, branch off a draft. - Note chunking
See exactly how your note is chunked for retrieval — token counts, overlap, boundaries. - Backlinks
Every note knows which other notes reference it. Build a personal wiki without effort. - Knowledge Base
Build a searchable, versioned KB from docs, PDFs, or scraped sources. Indexed automatically. - KB embed widget
Drop the chat widget on any site. Customers get instant answers from your KB. - KB settings
Tune chunk size, overlap, refresh schedule. Same controls as your Maestro pipelines. - Support tickets
When the KB can't answer, escalate to a human. Built-in ticket queue + AI triage. - Rooms
Group chat with humans and agents. Mix #channels and DMs in one workspace. - Room chat
Threads, mentions, members, files. Maestro can be added as a participant. - Email
A real email client built into the platform — many accounts, agents can read and reply. - Dashboard
A configurable widget grid — Maestro prompt, stats, charts, recent activity. Your morning view. - Workstation
Open many surfaces at once — chat + notes + browser side-by-side. Tiled, draggable, saveable. - Browser
Built-in web browser. Maestro can read a page, extract content, or hand it to a tool. - News reader
Pull every feed into one quiet reader. Summaries generated on demand. - Portfolio
A workspace gallery to show off finished work — clips, designs, demos, case studies. - Knowledge graph
Notes, people, projects, dates — auto-extracted into a graph you can browse and query. - osIndex
Search every local file with three scorers — lexical, semantic, recency. Stays on-device. - Code App
Repos, web IDE, pull requests and deploys in one surface. Pair-program with Maestro on the same files. - Repos list
Every repo in your workspace at a glance — public / private, last commit, deployment state. - README
A rendered README front-and-center — the first thing you see when opening a repo. - Repo chat
A Maestro scoped to one repo. Knows your code, files, and history without prompting. - Web IDE
A full Monaco workspace — multi-file editing, search, terminal, debugger. Works anywhere. - Manifest
A structured view of package.json — dependencies, scripts, versions in one place. - Docs
Auto-generated API docs from your code — components, hooks, functions, examples. - Env keys
Per-repo environment variables. Masked, audited, never in source control. - Issues
Lightweight issue tracking — labels, assignees, comments. Agents can file and close them. - History
Commit graph with messages, authors, durations. Click any commit to see the diff. - Versions
Switch between tagged versions of a repo without leaving the page. - Releases
Cut a release with notes, artefacts and changelog — auto-generated by Maestro if you want. - Merge
Compare any two branches with a full diff. Approve or request changes inline. - Extensions
Per-repo IDE extensions — linters, formatters, language servers. Click to toggle. - Deployments
Every build, every deploy, every rollback. Status pulses, logs stream live. - Repo members
Per-repo access — owner, maintainer, contributor, viewer. - Repo settings
Default branch, visibility, merge strategy, issue templates — one screen. - Pull requests
A workspace-wide PR dashboard. CI status, reviewers, conflicts — everything at a glance. - PR review
File diff and threaded comments. Maestro can summarise large PRs or suggest reviewers. - Import repo
Pull any Git repo into the workspace. Optionally keep it synced with the upstream. - Snippets
Your library of reusable code blocks — hooks, helpers, middlewares. Searchable and copyable. - Containers
Long-running containers for backend services, agents, or build workers. Logs stream live. - Build artefacts
Every build produces signed artefacts — versioned, downloadable, retained for as long as you need. - Records
A structured table view for any data model — customers, invoices, leads. Like a database with eyes. - Record detail
Each row gets a full form view with related records, audit trail and inline editing. - CSV migration
Drop a CSV, map columns, import. Maestro can suggest mappings automatically. - Data sources
Sync from Stripe, Postgres, Mixpanel, anything with an API. Channels keep your data fresh. - Add data source
Pick a connector or a webhook. Authenticate, map schema, sync. - Channel detail
Inspect schema, sync history and recent rows for any connected source. - Connectors
A catalog of every integration the platform supports — first-party and community. - Connections
See which integrations are live, their auth state and last sync. - Folders
A hierarchy of notes, files and KBs. Permissions cascade. Local + cloud per folder. - Folder access
Per-folder visibility — private, workspace, or shared with specific people and roles. - Folder chunking
Set chunking strategy per folder. Different content, different retrieval rules. - Memory groups
Group memories by customer, project or topic. Surfaces share or scope them as needed. - App databases
Each app gets its own Postgres database. Browse schema → tables → rows from the workspace. - Domains
Buy, attach and route custom domains in seconds. Wildcards and SSL handled. - Domain detail
DNS records, SSL status, routing rules — everything for one domain in one screen. - osStudio config
Every prompt, pipeline, agent, tool and memory group is a versioned config. Customize without source code. - Pipeline stages
Define and version every stage of your AI pipeline — retrieve, rerank, generate, post-process. - Retrieval stages
Chain retrieval stages into a custom pipeline per path scope. Override the defaults at any level. - Agent configs
Edit any agent's name, model, tools, prompt and policy as a versioned record. Fork, publish, share. - Tool configs
Define new tools as inputs + handler. Anything you can call from Maestro. - Skills
Skills are markdown playbooks Maestro reads when conditions match. Versioned per workspace. - Prompts
Every system prompt is a versioned record. Diff between versions, roll back, share to community. - Connector configs
Authoritative source for every connection — scopes, secrets, OAuth config. - Memory groups (Studio)
Wire surfaces into memory groups with retention, write mode and PII redaction policies. - Plugin · retrieval
Drop a JS function into the retrieval pipeline — rerank, dedupe, boost recent, anything you can code. - Plugin · post-hook
React to chat events after the fact — audit log, save artefact, fire webhook. - Plugin · routing rule
Route messages to a specific model or agent based on tags, content, context. - Plugin · guard
Block or escalate any tool call with a small JS rule. Per-workspace policy as code. - Plugin · command
Add slash-commands to Maestro — your own /weekly, /summary, /standup. - Plugin · tool
Write a tool in JS, ship it as a versioned config. Maestro picks it up immediately. - Tool UI plugin
Custom React components that render tool results — tables, charts, anything you need. - Community browse
Install community-shared agents, prompts, tools and plugins. Star, fork, contribute back. - Versions
Branch any note, KB or repo. Compare versions side-by-side. Merge or roll back. - Merge Station
A 3-way merge editor — mine, theirs, merged — for any branched entity in the workspace. - Sharing
Share any surface with a link or with specific people. Granular per-resource permissions. - Members
Invite teammates with roles — owner, admin, editor, viewer. Custom roles supported. - Approvals
Risky agent or tool calls queue here for a human to approve. Audit trail attached. - Audit log
Who did what, when. Filter by user, action, resource. Export for compliance. - Live presence
See cursors and selections of teammates editing the same surface in real time. - Sandbox
Define default execution mode and auto-approve / block lists per workspace. - Active sandboxes
Every running sandbox with its task, age and active resources. Kill or extend on demand. - Ingestion
Workspace-wide chunking defaults and skip-patterns. Stops indexing junk. - Model access
Manage every BYOK key from one screen. Provider status, cost caps, rate limits. - Integrations
OAuth-connected services — Slack, GitHub, Stripe, Google. Auth state in one view. - Usage
Tokens, runs, dollars — every cost, by category, in a chart you can drill into. - Outbound webhooks
Fan out platform events to your own endpoints. Auto-retry, signature verification. - Erasure
GDPR-style data deletion with retention policies for audit, chunks and memories. - Secret rotation
Schedule automatic rotation for every secret. Manual override per key. - Secrets overview
All secrets, all environments, one screen. Health, age, leaked-credential alerts. - Security audit
Every login, every privileged action — searchable, exportable. - Security alerts
Real-time alerts for unusual logins, secret leaks, anomalous spend. - Alert rules
Define thresholds in plain language — cost, error rate, login geography, secret expiry. - API tokens
Issue scoped tokens for CI, mobile, analytics. Mask, rotate, expire. - Changelogs
Every platform release with concise notes — see what shipped, when. - System logs
A filterable feed of system events — info, warning, error — across all your apps. - Custom domains
Attach your domains. CNAMEs verified, SSL auto-issued, downtime tracked. - Container limits
Per-workspace CPU, memory and runtime quotas. Live consumption visible. - Infrastructure
Run osFoundry in your own cloud — VPC, database, storage all yours. - Command palette
Find anything in two keystrokes. Pages, actions, recent items — fuzzy ranked. - Shortcuts
Customizable keyboard shortcuts for every common action. - Screen record
Quick clips for bug reports, demos and walkthroughs. Saved to the workspace. - Translate
Translate any selection on the fly — into 100+ languages, with your BYOK model. - Notifications
A single inbox for everything the platform needs to tell you. Filter by source. - Workspace switcher
Hop between personal, team and partner workspaces. Each one fully isolated. - Favorites
Pin the surfaces you visit most. Quick access from anywhere in the app. - Clipboard
Your last copies, kept for as long as you need. Re-paste with one click. - Create new
A context-aware "new" menu — note, page, repo, agent, anything — with shortcut hints. - Ship Room Apps internally
Build apps that run inside your workspace — no public URL, no DNS, no deploy. Just invite teammates. - Publish Room Apps publicly
Flip a switch to go public. Branded subdomain, edge cache, SSL — without leaving osFoundry. - Per-app Postgres
Every app gets its own Postgres with pgvector and TimescaleDB. Pay only for what you store. - Per-app KV store
Redis-compatible KV with TTLs. Sessions, caches, counters — sized to your app. - Per-app file storage
S3-compatible blob storage per app. Versioned, signed-URL access, no bucket sprawl. - Per-app secrets vault
Encrypted secrets with scheduled rotation. Masked at rest, audited on every read. - Serverless functions
Lambda-style functions per app — cold start in 40ms, no provisioning, billed per millisecond. - Container compute
Need persistent processes? Flip the runtime to a container. Same app, different mode. - Native runtime
On desktop, your app can run as a native process with filesystem and shell access — gated by a permission dialog. - App cron jobs
Schedule recurring tasks per app. Cron syntax, retries, observable runs. - App version history
Every deploy is a versioned snapshot. Rollback in one click — code, schema, secrets all together. - Per-app subdomains
Each app gets `<slug>.apps.osfoundry.io`. SSL automatic. Bring your own domain too. - Cross-app data grants
Apps can share data with other apps — explicit scoped grants, audited, revocable. - Public app rate limiting
Tunable per-IP and per-token rate limits on public apps. Defaults that handle the hug of death. - Visitor authentication
Anonymous, magic-link, OAuth or workspace SSO for public apps. Mix and match per route. - Offline-capable apps
Apps keep working offline. Writes queue locally and replay on reconnect. - Build pipeline & CI
Every push triggers install → lint → build → deploy. Logs streamed, artefacts retained. - Public KB hosting
Your KB on your own domain, fully indexed, with a chat widget and ticket queue out of the box. - Real-time data channels
Pub/sub channels with at-least-once delivery, history replay and per-subscriber cursors. - Change-data capture
Stream every row change out of any app database. Wire CDC into agents, channels or webhooks. - Inbound webhooks
A URL per app or workspace for receiving events. Signature verification baked in. - Hot replay & history
Every channel keeps the last N days of events. Reset a cursor, replay from any seq id. - MCP server fleet
Install MCP servers per workspace. Tools from GitHub, Figma, Postgres — addressable from any agent. - Local folders
Mount real folders from disk. Files stay local; only chunks travel to the cloud, if you let them. - 250K+ public datasets
Browse and import from a quarter-million public datasets — pre-indexed, ready for fine-tuning. - Bring your own datasets
Upload CSVs, JSONLs or stream from your warehouse. Same indexing pipeline as built-in data. - Local sync engine
A bi-directional sync engine for notes, chats, KBs and chunks. Reconciles on reconnect. - Realtime sync across devices
Edits propagate to every device within a couple hundred milliseconds — laptop, phone, tablet. - Workspace-wide search
A single search across notes, code, KBs, chats, records and memories. - Local-first mode
Your data lives on your device. Cloud is opt-in per surface. - Private workspace
No telemetry, no cloud upload, no third-party LLM calls. BYOK only. - No-leak LLM mode
All requests route to local models. Cloud providers blocked at the network layer. - End-to-end encryption
Workspace keys never leave your devices. Server sees ciphertext only. - Enterprise SSO
SAML and OIDC against Okta, Azure AD, Google Workspace or any IdP. SCIM provisioning included. - RBAC & permissions
Granular roles per workspace and per resource. Custom roles supported. - Anomaly detection
Spend, error rate and login anomalies trigger alerts automatically. Tunable thresholds. - Workspace policies
Codified policies for tools, regions and approvals — versioned, enforced workspace-wide. - Data residency
Pin your workspace to a region. Multiple regions supported with replication and failover. - Pure usage billing
No seats, no plans, no commitments. Just the resources you use, priced clearly. - Model quantization
Quantize any model down to q4 / q3 in a few clicks. Run a 70B model on a laptop GPU. - Hot-swap LoRA
Swap LoRA adapters on a running base model in under 50ms. Different personas, same weights. - Schedule agents 24/7
Agents that run on a cadence — daily briefs, hourly checks, weekly reports. They wake themselves up. - Tool scope per agent
Each agent gets a tool allow-list. Researcher can read the web; coder can execute commands. - Agent task queue
Long-running and recurring work lives in a prioritised queue. Survives restarts, replays on failure. - Memory extraction
Maestro extracts durable memories from every conversation — facts, decisions, preferences. - RAG with reranking
Hybrid lexical + vector retrieval, reranked with a cross-encoder. State-of-the-art recall. - MMR diversity
Maximal Marginal Relevance picks results that cover different angles, not just the top match. - Auto-chunking pipeline
Smart chunking that respects sections, code blocks and lists. Tunable per folder. - Markdown config
Author any config as Markdown with frontmatter. osStudio reads it as a versioned record. - Paper
Write academic papers with passage-linked thinking spaces and research beside the manuscript. - Memo
Capture quick ideas in folders with a focused editor and a choice of synced or local storage. - ADE
Give each coding task its own worktree, agents, and review loop. - GitNest
Bring repositories, work, and code reviews into one decision queue. - Game
Create game projects, review AI proposals, and organize playtests.
Explore by capability
- Run any AI model with osFoundry — local, cloud, or self-hosted
BYOK any cloud API, run open weights on your laptop, or deploy a dedicated GPU endpoint — all from one workspace. - Train and fine-tune AI models on osFoundry
Fine-tune Llama, Mistral, or Qwen with LoRA on your data. Quantise for cheap inference. Hot-swap adapters at runtime. - Build AI agents and automations with osFoundry
Persistent agents that schedule, plan, use tools, share memory, and run 24/7 — without you watching. - Build internal tools and public apps with osFoundry
Ship a tool, hit Deploy, get a public URL. Database, secrets, cron, logs all included. Replaces team chat, CRM, and dashboard tools. - Knowledge bases, memory, and RAG on osFoundry
Auto-extracted entities, knowledge graphs, custom retrieval pipelines, and reranked RAG — all configurable in osStudio. - Code repos, AI pair programming, and PR reviews on osFoundry
Web IDE, GitHub import, PR reviews, and Maestro chat across your repos — without leaving the workspace. - Realtime data, webhooks, OAuth connectors, and MCP on osFoundry
Native realtime channels, change-data capture, hosted OAuth, and a fleet of MCP servers — all built into the workspace. - Built-in browser, web scraping, news firehose, and email on osFoundry
AI agents that browse logged-in sites, scrape pages on a schedule, and ingest from 2,000+ news sources. - Customise Maestro and your AI stack with osStudio
Edit every prompt, pipeline, agent, retrieval stage, routing rule, and tool as a versioned config. Fork, branch, share. - Local-first, private, and offline-capable AI with osFoundry
Run models on-device, keep private channels off the cloud, and work offline. Your data, your machine. - Enterprise security, SSO, audit, RBAC, and BYO-VPC on osFoundry
WorkOS SSO, audit logging, RBAC, anomaly detection, secret rotation, GDPR data erasure, multi-region residency, and BYO-cloud. - Pure-usage pricing, public datasets, sites, and custom domains on osFoundry
Pay only for what you use. 250K open datasets indexed. Page builder, website builder, and public KB hosting included.