Aisha Khan9:38 PM
Release window is 9pm. Are we actually good to go?
Pay once. Unlimited users on your server.
Self-hosted · Open-source frontend
An agent can only do what the person behind it could.
Checked live on every call, and written to the log before it acts. Chat, docs, tasks, video and calendar come with it, in one Docker deploy, with no per-seat fees and nothing leaving your network.
- 09:14:02task.createallowed#engineering · “Cut 2.7.0 release notes”
- 09:14:02doc.readallowedRelease checklist
- 09:14:03channel.postrefused#financePriya cannot post in #finance, so neither can her agent
Each entry hashes the one before it. Written before the action, so a refusal is on the record too.
Acme Inc
Search workspace…
AK
engineering
Release Captain · AI teammate · in this channel
Message #engineering…
Preview of OneCamp at /app/channel/engineering
Product tour
See it before you commit
Walk through the real app: channels, AI, docs, tasks, and more. No slick mockups.
ChannelsAIDocsTasksVideoCalendar
Full product walkthrough · real app, no mockups
Governed AI
Permissions, not promises
Most AI workspaces ask you to trust that the assistant will behave. This one is built so it cannot misbehave quietly.
An agent can only do what its author could
Every action is checked against the live permission graph of the human who authorised it: channel, project, and document membership, read at the moment of the call rather than cached at setup. Remove someone from a channel and the agents acting on their behalf lose it on the next request, not at the next token rotation.
If it can't be recorded, it doesn't happen
When an agent calls a tool, the audit entry is written before the call and a failure to write it refuses the call. That ordering is the point: a decision that was made and never recorded is worse than one recorded and abandoned, because only the second is discoverable afterwards. Administrative changes are logged too, off the request path, so a slow write never blocks an admin.
Refusals are logged too
A denied call leaves a row with the reason it was denied, the credential, the named agent, and the human behind it. An access review can therefore answer what an agent tried and was stopped from doing, which is the question that actually matters after an incident.
History that can't be quietly edited
Each audit entry hashes its own contents plus the previous entry's hash. Any later insertion, edit, or deletion breaks the chain, and a verify endpoint recomputes it and reports the first divergence. Exports carry the per-row hashes so an auditor can check them without trusting the UI.
The model can stay in the building
Local inference through Ollama by default. Turn on local-only mode and the server refuses to activate a cloud provider at all, rather than warning you and allowing it. Point it at OpenAI or Anthropic when you want to, with PII redaction on the way out.
So can the meeting audio
Speech to text runs on your server too, from a Whisper model in your own compose stack, so a recorded call is never uploaded to a transcription vendor and nothing is billed per minute. Deepgram and Google are still there if you prefer them. The difference is that the settings page tells you which option sends your meetings where, instead of leaving you to work it out from a vendor's name.
Offboarding reaches the agents
Deactivating someone is re-evaluated live at every authorisation surface, so their agents and API tokens stop working immediately without anyone hunting for credentials to revoke. Bots and attribution-only identities can never be the authority for a call in the first place.
Enterprise controls
The boxes procurement makes you tick
Already in the box. No enterprise tier, no add-on SKU.
Identity
- SAML 2.0 single sign-on
- OIDC, for anything modern
- LDAP / Active Directory
- Two-factor auth with recovery codes
Lifecycle
- SCIM 2.0 provisioning
- Automatic deprovisioning on offboard
- IdP-managed accounts can't add a local password
- Role and project permission model
Evidence
- Tamper-evident admin audit log
- Chain verification endpoint
- CSV and JSON export with row hashes
- Named actor: human, integration, or agent
Data
- Runs entirely on your infrastructure
- Residency follows your server
- Local-only AI mode
- PII redaction before any outbound call
Why we built it
AI got access before anyone agreed to it
Not a compliance problem. A Tuesday morning problem.
An assistant gets wired into the wiki, the tickets, and the chat, and within a week it can reach more than most of the people who work there. Nobody decided that. Ask who authorised a particular action and the honest answer is usually a shrug and a log line with a token id in it. OneCamp starts from the other end: an agent inherits one person's permissions, re-checked on every call, and nothing it does happens without a record naming who stood behind it. OneMana runs the company on this, and the product still changes every week.
OneCamp
all in one
SlackChat
NotionDocs
AsanaTasks
ZoomVideo
CalendarSchedule
What buyers say
I purchased OneCamp, tried it out, and still use it. The quick video chat works, chat between users works, I can make multiple files to share with anyone, and it's easy to invite a colleague with an email request. You can easily ask the AI box questions, which is useful. It's truly an all-in-one build. Worth it, with a responsive, friendly developer. I'm satisfied.
Something really sweet if you like to self-host for your team. Your team's self-hosted command center: chat, tasks, docs, and video meetings, all in one workspace you own. No per-seat pricing. No AI credits. No vendor lock-in. Just install, invite your team, and start collaborating. One payment, and it replaces four subscriptions.
Runs in production
Same app OneMana ships from
Live demo
Open-source frontend
0
agent actions that run unaudited
∞
people, no seat tax
0%
on your infrastructure
<0 min
to get running
What's inside
The work the AI is governed over
Governance is only worth something if there is real work behind it. This is the surface the agents operate on, under one login on one server.
- AI agents
- Build a teammate, not just a chatbot. @mention it in a channel or DM and it answers in-thread as its own badged member, or let it run on a schedule to triage, file tasks, and post recaps. You choose whether it acts on its own or proposes and waits for your OK. It can only ever do what its owner could, every call is checked live, and every call is recorded before it runs.
- Local AI
- Runs locally via Ollama by default, so chats and docs never leave your box. Point it at OpenAI, Anthropic, or your own endpoint when you want; local-only mode and built-in PII redaction keep you in control. Ask once and get a cited answer from your channels, docs, tasks, and connected Gmail or GitHub. Daily token budgets included.
- Audit trail
- Every admin change and every agent tool call, allowed or refused, with the reason and the human accountable for it. Hash-chained so edits to history are detectable, verifiable on demand, and exportable as CSV or JSON with the row hashes intact.
- SSO and provisioning
- SAML, OIDC, and LDAP for sign-in, SCIM 2.0 for joiners and leavers, TOTP two-factor with recovery codes for everyone else. Directory-provisioned accounts authenticate at your IdP and cannot be given a local password that routes around it.
- Tables
- Notion-style databases with grid, board, calendar, and chart views. Typed columns, rows that link to real tasks and projects, and an AI that builds the whole table from one sentence, fills a column row by row, or answers a question straight from your data with a chart. Replaces Airtable.
- Automations
- Slack-class "when this, do that" without the bot zoo. New message with 'bug:' in #support, auto-file a triage task. Describe the rule in plain English and it runs on your server, forever.
- Chat
- Channels, threads, DMs, reactions, files. Real-time over MQTT. We built this first and still dogfood it daily at OneMana.
- Tasks
- Kanban in the same app as your channels. No Zapier keeping Asana in sync with Slack.
- Docs
- Block editor, slash commands, live cursors when two people land on the same page. Good enough that we cancelled Notion.
- Whiteboard
- An infinite canvas for diagrams, flows, and mind maps with live cursors. Ask the AI to draft a flowchart or UI mockup, then edit it together. Replaces Miro and FigJam.
- Video
- LiveKit runs on your hardware. Calls, screen share, recordings, and an AI recap posted to the channel after the call. Transcription stays local too.
- Calendar
- Google sync works both ways. Task due dates show on the team calendar, and the AI finds a meeting time and preps you before it. Without another tool.
- Teams
- Roles, projects, the usual permission model. The AI only sees what the logged-in user can see, and so does anything acting on their behalf.
- Programmable
- A scoped public API, an official TypeScript SDK, and an MCP server so any AI client can drive your workspace, plus shareable templates for agents, automations, and tables. Tokens never exceed what their owner could do by hand, and calls made through them are audited like any other.
- Your server
- One Docker command, SSL included, open-source frontend on GitHub. No phone home, no per-seat invoice, no vendor lock-in. Share a doc, board, or table as a read-only link with people outside your org when you need to.
Inside the product
What it actually looks like
Ask the AI to brainstorm a mind map on the whiteboard, or build a table from a sentence. Calendars with a live now-line. Docs with someone else's cursor in them. The same UI we use at OneMana.
Acme Inc
Search workspace…
AK
AI
DCAK
Generate with AI
AutoFlowchartRoadmapMind mapOrg chart
Describe what you want to draw…Generate
Preview of OneCamp at /app/board/strategy
Getting started
One person installs. Everyone else logs in.
Most teams are up in under ten minutes. SSL, database, and AI models are handled by the installer.
Run one command
SSH into any Docker-capable server and run the installer. It pulls images, wires SSL, and sets up the database.
Connect your directory
Point SAML, OIDC, or LDAP at it and switch on SCIM so joiners and leavers handle themselves. Or just send email invites and skip this.
Give the AI a job
Build an agent, decide whether it acts or asks first, and watch what it does in the audit log. It inherits the permissions of whoever authorised it.
Works on every device. Web, PWA with push notifications, no app store required.
Switching
The two things that actually stop teams moving
Neither of them is price. You would lose your history, and somebody has to run it. Here are both answers, including the parts that are not flattering.
Bring your Slack history with you
Export your workspace from Slack, upload the archive, and OneCamp imports it.
- Channels, messages and threads, in order
- Direct messages and group conversations
- Files, and the reactions on everything
- People matched to their new accounts, avatars included
It shows you a plan before it writes anything, de-duplicates so a second attempt cannot double-post your history, and can be rolled back if you change your mind.
What it will not bring
- Microsoft Teams exports, not yet
- Slack apps, workflows and integrations
- Huddles, and anything Slack does not put in an export
Said here rather than discovered afterwards. An import that oversells itself costs more trust than the features it was hiding were worth.
And someone has to run it
That is the real question about self-hosting, so here is the honest version of the answer.
- Install
- One command. SSL, database and models included, usually under ten minutes.
- Updates
- One command, and it refuses to start against a schema it does not match rather than corrupting anything.
- Backups
- Scheduled at install and watched like storage, with an honest warning that a backup on the same machine is an undo buffer, not disaster recovery.
- Restore
- Brings the schema forward, restarts in place, and never leaves the application stopped.
- Health
- One command tells you what a new workspace still needs, including whether it can actually send email.
It is still your server. If nobody on the team wants that job, a subscription is the right answer, and we would rather you heard that here.
Pricing
Buy once, or let us host it
One lifetime license for self-hosting, or fully managed OneCamp Cloud. Cloud includes a self-host license if you ever want to move.
The math
The AI in those tools is usually a per-seat add-on on top of the seat you already bought. OneCamp includes the agents, the local AI, the SSO and the audit log in one install, with unlimited users. Put your own team size in below.
Pay once. Own it forever.
What teams pay today
Five separate subscriptions
- Stack · Slack, Notion, Asana, Zoom, calendar
- Annual spend · Per seat, per tool, growing with headcount
- AI · Per-seat add-on, in someone else's cloud
- Meetings · Transcribed by a vendor, billed per minute
- Billing · Per seat, per tool, every year
With OneCamp
One workspace, flat pricing
- Stack · Chat, docs, tasks, tables, video, calendar, AI
- AI · Included, governed, runs on your hardware
- Meetings · Transcribed on your server, no per-minute bill
- Billing · Unlimited users, no per-seat fees
- Choice · Lifetime self-host or managed cloud
Self-Host · Lifetime
Popular$99once
₹9,999 · INR · taxes included
- Unlimited users. No per-seat fees
- All modules incl. local AI and agents
- SSO, SCIM, MFA, and the audit log included
- Open-source frontend, yours forever
- Runs on your own server
- Free updates within your major version
OneCamp Cloud · Managed
$99/ month
₹9,999/mo · up to 30 users
- Fully managed hosting on your own subdomain
- Your own server. No database shared with anyone
- We handle SSL, monitoring, and uptime
- AI teammates included. Bring your own model key
- Includes a self-host license. Switch anytime
- We set everything up within 12 hours
people.
Slack + Notion + Zoom
$10,020
per year, and it grows with the team
- Slack Pro, $9/user/mo
- Notion Business, $10/user/mo
- Zoom Pro, $13/user/mo
- Notion AI, at its old add-on price, $10/user/mo
OneCamp, self-hosted
$243
first year, and it does not grow with the team
- $99 licence, paid once, unlimited users
- $12/mo server that runs it
- AI included: local models, or your own API key billed at cost
- Every year after this one is just the server
At 20 people that is 41× less, and the gap widens with every person you add, because one side charges per seat and the other does not.
What this does not count: someone has to run the server. Install is one command and updates, backups and restore each have one too, but it is still your machine. If nobody on the team wants that job, a subscription is the honest answer and we would rather say so here than after you have paid.
Hardware
Rough sizing
Starting points from real droplets, not lab benchmarks. Your mileage varies if everyone is on video calls while the AI summarizes docs.
Under ~50 people
8 GB RAM · 4 vCPU
What we run OneMana on. Chat, docs, tasks, and CPU AI are fine here.
50–200 people
16 GB RAM · 4+ CPU
Add headroom if local AI is on all day or video is heavy.
200+ people
32 GB RAM and up
Treat this as a conversation, not a formula. We'll help you size it.
Questions
Before you buy
The emails we actually get. Short answers.
Can the AI see things it shouldn't?
It checks your channel, project, and document membership at the moment of the call, so it can only read what you can already open. The same check governs anything acting on your behalf: an agent, an API token, an external MCP client. Nothing is cached at setup time, so a permission you lose is a permission it loses on the next request.
Is there an audit trail for what the AI did?
Yes, and refusals are in it too. Each row names the tool, the decision, the reason, the credential, the agent if there was one, and the human accountable. The entry is written before the action runs, and if it cannot be written the action does not happen. Entries are hash-chained so later edits to history are detectable, and you can verify the chain or export it with the hashes.
What happens when someone leaves?
Deactivate them in OneCamp, or let SCIM do it from your directory. Eligibility is re-checked at every authorisation surface on every call, so their sessions, API tokens, and any agent acting on their authority stop working immediately. You don't have to go looking for credentials to revoke.
Do you support SSO and SCIM?
SAML 2.0, OIDC, and LDAP for sign-in; SCIM 2.0 for provisioning and deprovisioning; TOTP two-factor with recovery codes for accounts that sign in with a password. Accounts your directory creates must authenticate at your directory, so nobody can give them a local password that bypasses it.
Where does the AI run?
On infrastructure you choose, never ours. Point it at OpenAI, Anthropic, or any OpenAI-compatible endpoint, which covers vLLM, LM Studio, OpenRouter, Groq, and a gateway you run yourself. Local models through Ollama are a switch you turn on, so nothing pulls a model down unless you ask for it. Turn on local-only mode and the server refuses to activate a cloud provider at all rather than warning you and letting it through, and PII redaction runs before anything outbound.
Does OneCamp Cloud include AI?
The AI edition is what gets provisioned, so the agents and the teammates are all there waiting. The model is the part you bring: add a key from OpenAI or Anthropic, or point it at any OpenAI-compatible endpoint. I do not resell inference and I never see your prompts, so you pay your provider directly at their price. Until you add a key the AI surfaces sit idle, and everything else in the workspace works exactly the same.
I pay once and that's it?
Yes. One license key, unlimited users, no annual renewal. The AI, the agents, the SSO, and the audit log are all included. None of it is an enterprise tier you unlock later.
Can I share a doc/board with someone outside the org?
Yes. Share a doc, board, or table as a read-only link with people outside your team, and create guest links for calls. Guests see only what you shared, nothing else, and every open is audited for you.
Do I need a GPU?
No. Most people point OneCamp at a provider key, and then the model runs on the provider's hardware rather than yours. If you switch on local models, the small default is fine on CPU with 16 GB RAM, and a GPU is what makes the bigger ones practical.
Can we import Slack history?
There's a Slack import built in. Bring channels and messages over when you're ready to switch, not before.
Does everyone need to install something?
One person runs the installer on a server you control. Everyone else opens a browser or adds the PWA to their phone.
What about phones?
Mobile web and PWA work well, push included. We don't ship App Store builds. Most teams don't miss them.
Install broke. What now?
Open a GitHub issue with the log. Most installs finish in under ten minutes on a normal VPS. We'll help you untangle it.
What if OneMana shuts down?
You already have the backend binary and an open-source frontend. Your instance does not phone home. It keeps running.
If you can't say what your AI is allowed to do, this is for you.
Bounded by your permissions. Audited before it acts. On hardware you own.
See pricing for lifetime and cloud plans
Instant license key · GST invoice · Unlimited users