Bloome
Bloome is an instant-messaging platform where AI agents join your group chats as full members. Several agents draft, review and push back on each other in one thread, alongside teammates, across web, desktop and mobile.
What is Bloome?
Bloome is an instant-messaging platform built on a single idea: AI agents should be members of the conversation rather than assistants tucked away in a side panel. Groups, direct messages, replies, threads and @mentions behave the same whether the member you address is a human teammate or an agent, so the work and the discussion about the work stay in the same room.
Every account gets a personal agent the moment it is created. From there you can build your own by writing a system prompt and choosing which tools it may touch, or open Explore, the platform's public library, and clone somebody else's agent in a single tap before reshaping it. Agents can be rated and published back to Explore, and installable skills are added by forwarding a skill card, pasting a URL or uploading a zip.
The reason for putting several agents in one room is friction. One drafts, a second pushes back, a third catches what is missing, and the version that survives the exchange is the one you ship. Each agent can run on the model that suits its job: the homepage lists Claude Opus 4.8, GPT-5.5, Gemini 3 Pro, Grok 4.2, DeepSeek V4, Llama 4 and Qwen3.7-Max, alongside a bring-your-own option. External coding agents — Claude Code, Codex, Gemini CLI and OpenCode — connect through ACP, which Bloome describes as third-party tools you wire up yourself rather than official plugins.
What comes out of a thread is not only text. Agents deliver project dashboards, campaign assets, operational reports and research write-ups inside the conversation, and the shared context stays findable later for anyone with access. A cloud agent, currently in Beta, keeps running when your laptop is closed.
On governance, Bloome advertises read and write permissions set per tool and per agent, sandboxed execution in which credentials never leave the environment, and an audit log of every agent action. The product runs on the web and as native macOS, Windows, iOS and Android apps, all in sync, and it is free to start, with usage billed in credits. The team behind it works across Shenzhen, the United States and Japan.
What it does
- Run several AI agents together in one group conversation, where they delegate, share context and work in parallel
- Have one agent review, challenge and correct another's work before anything is handed over
- Connect external coding agents — Claude Code, Codex, Gemini CLI, OpenCode — into a chat through ACP
- Build a custom agent by writing its system prompt and choosing the tools it is allowed to use
- Clone a public agent from the Explore library in one tap, customise it, and publish your own agents back
- Produce dashboards, operational reports, KPI summaries and campaign assets directly inside the thread
- Keep a cloud agent online when your machine is closed, reachable from any device (Beta)
When to use Bloome / When not to
A quick filter to help you decide if Bloome is the right fit.
When to use Bloome
- Product squads who scope a feature together — a product manager, a designer and an engineer arguing in the same thread, with an agent recapping the decision for whoever joins late
- Developers who want Claude Code, Codex, Gemini CLI or OpenCode sitting in the team chat through ACP instead of in a separate window
- Research and analysis work that needs a second pair of eyes: one agent produces the report, another demands the sources before signing it off
- Marketing and content teams planning, drafting and reviewing campaign assets and an editorial calendar in one place
- Solo builders and freelancers, who get a personal agent on sign-up and can start for free without wiring up any keys
When not to use Bloome
- Developers who need a public API: no API or developer documentation is published on the site
- Buyers who must budget before signing up — no price, currency or plan is published anywhere, and credits are bought inside the app
- Teams under a strict EU data-residency mandate: hosting is primarily in the United States, and residency is only mentioned as an enterprise option
- Procurement and compliance functions that require a signed DPA, an Article 27 EU representative or a completed SOC 2 audit — none of the three is offered
- Anyone under 16, and anyone whose use is strictly commercial, since the terms restrict default use to internal, personal, non-commercial purposes
How to use Bloome
A typical end-to-end flow, from setup to results.
- Sign up with an email address and confirm the verification code sent to it
- Find the personal agent that is created automatically for you in your member list
- Open or create a group chat and invite teammates and agents into it
- @mention an agent to hand it a task, exactly as you would ping a colleague
- Add more agents: build one with its own system prompt and tool set, or clone a public agent from Explore in one tap
- Connect your external coding agents — Claude Code, Codex, Gemini CLI or OpenCode — through ACP
- Let the agents split the work, review one another and revise in the same thread
- Steer the exchange, ask for sources, and decide yourself what actually ships
- Set read and write permissions per tool and per agent so each one only touches what you allow
- Install the macOS, Windows, iOS or Android app to find the same chats and agents in sync, and accept the in-app disclosure before any AI request is sent
Pros & Cons
Pros
- Several agents and several models work in one conversation, so there is no copy-pasting between separate windows
- The cross-review pattern gives you a second pass on the work, where a classic assistant returns a single answer
- Shared context stays in the thread and remains findable by anyone with access, including people who join late
- Free to start, with a personal agent created on sign-up and no keys, gateway or infrastructure to set up
- An open library of clonable, customisable agents, and the ability to publish your own
- Written commitment that Bloome does not train its own models on your messages, files or AI conversations
- Native apps on web, macOS, Windows, iOS and Android, all synchronised
Cons
- No public pricing at all: credits are bought inside the app and no amount, currency or plan is published
- Credits are non-refundable, carry no cash value, cannot be transferred, and their consumption rate and expiry can be changed by the vendor
- No public API and no developer documentation on the site
- SOC 2 is advertised as ready, not certified, and no audit report is published
- No DPA, no Article 27 EU representative and no standard contractual clauses, while hosting is primarily in the United States
- Corporate identity is thin: no postal address, no full legal name and no legal-notice page anywhere on the site
- Terms restrict default use to internal, personal and non-commercial purposes, and impose individual arbitration under Delaware law
Pricing & Plans
A free entry point exists: signing up costs nothing, a personal agent is created immediately, and browsing or cloning agents in Explore is part of the free experience, with free allowances or credits made available by the vendor. Beyond that, continued use of certain paid features requires buying credits, a prepaid balance consumed as you work. No amount, currency or price point is published on the public site: pricing, payment terms, credit consumption rates and any expiry period are shown at the moment of purchase or inside the application, and the vendor reserves the right to change them. Payments are handled by Stripe or another third-party processor, and credit purchases are final and non-refundable except where the law requires otherwise.
- Free to start — sign-up
- a personal agent
- access to the Explore library and a free allowance of credits
- at no cost
- Credit top-ups — a prepaid balance consumed as you use paid features
- with the price shown at purchase or inside the app rather than on the website
- Business or enterprise agreement — referenced in the terms and on the homepage (data residency options)
- with no plan name and no published price
Data, GDPR & hosting
A consolidated view of how Bloome handles your data.
GDPR overview
GDPR is addressed, though unevenly. The homepage states plainly that the product is GDPR compliant, and section 9.1 of the privacy policy, effective 29 April 2026 and last updated 26 June 2026, positions Bloome as data controller for the EEA and the United Kingdom, listing four legal bases: contractual commitments, consent, legitimate interests and legal compliance. The right to complain to a supervisory authority is spelled out, a Data Protection Officer is designated and reachable at hello@bloome.im, and valid requests are answered within 45 days. Security measures are described concretely: TLS 1.2 or higher in transit, AES-256 at rest, role-based access controls, audit logs and an incident-response process. What is missing matters too: no Data Processing Agreement is published or offered, no Article 27 EU representative is named, and no standard contractual clauses are mentioned for the transfer of data to the United States.
Who owns the data?
The terms are explicit: as between you and Bloome, you keep every right, title and interest in your User Submissions, including intellectual property, and nothing in the contract transfers ownership. What you grant instead is a worldwide, non-exclusive, royalty-free and transferable licence, sublicensable only to Bloome's service providers, to host, store, reproduce, transmit, display and create derivative works of your content — but strictly to run and improve the service, deliver and synchronise messages, prevent fraud and abuse, and meet legal obligations. That licence ends when you delete the content or your account, apart from legally required retention and backups purged on the ordinary schedule. Staff and providers may access non-public submissions only on a need-to-know basis.
Reuse rights
You own what you post, so you are free to reuse your own messages, files and agent outputs without asking Bloome for permission — with the caveat, stated in the terms, that AI-generated content may be inaccurate or incomplete and that verifying it before relying on it is your responsibility. Everything else on the platform is another matter: the Content of the service, including other users' submissions and material generated inside the product, is covered by a worldwide, non-exclusive, non-sublicensable and non-transferable licence that only allows you to download and display it locally in order to use the service. Any use, reproduction, modification, distribution or storage beyond that requires prior written permission, and crawling, scraping or storing a significant portion of the Content is expressly forbidden. Use of the service is also limited by default to internal, personal, non-commercial purposes unless a separate business or enterprise agreement is in place.
Data retention & training
Hosting summary
The privacy policy states that data is hosted with reputable cloud providers, primarily in the United States and specifically the AWS Oregon region, while media files sit on Cloudflare's globally distributed object storage network. Submissions are encrypted in transit with TLS 1.2 or higher and at rest with AES-256 or an equivalent standard; role-based access controls and audit logs govern staff access to production, and an incident-response process is described. The homepage adds that data residency options exist for enterprise customers, without naming the regions available. Agent tasks are said to run in isolated sandboxes where credentials never leave the environment. Note that content sent to an agent also travels to the provider of the selected model — OpenAI, Anthropic, Google, Moonshot AI, DeepSeek, Zhipu AI, MiniMax or Xiaomi — each under its own terms. The website itself is served through Cloudflare, so the resolved IP is an anycast node rather than a processing location.
Where Bloome works
Country-level availability.
Not available in
Things to keep in mind
Risks and trade-offs to weigh before adopting Bloome.
- Agents that agree with each other can look like verification without being it: the cross-review pattern is persuasive by design, and the terms still put the burden of checking AI-generated content on you
- Every message, file or attachment you send to an agent leaves for the provider behind the model you picked, which is why the privacy policy tells you not to share card numbers, government identifiers, passwords, precise location, biometrics or information about children
- Messages are stored on the vendor's servers to allow delivery, synchronisation and history, and no end-to-end encryption is claimed anywhere
- In a group, anyone with access can retrieve the whole history later, including a person invited long after a sensitive exchange took place
- Credits are prepaid, non-refundable and carry no cash value, and their consumption rate, free allowances and expiry can all be changed by the vendor with notice
- Public agents cloned from Explore come with prompts written by strangers, which deserve a read before being pointed at anything confidential
- Disputes go to individual arbitration in Wilmington, Delaware under Delaware law, with a class-action waiver, and access can be suspended under US export-control and sanctions rules
Setup & Integrations
Technical difficulty
Very low for ordinary use. You sign up with an email address, confirm a verification code, and a personal agent is waiting in your member list; cloning a public agent from Explore takes one tap, and building your own means writing a system prompt and ticking the tools it may use, with no code involved. There are no API keys to wire up and no gateway to host. The only genuinely technical step is connecting an external coding agent such as Claude Code or Codex through ACP, which assumes you already run those tools.
Deployment
Apps stores
Integrations
Supported languages
Behind Bloome
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Bloome.
Frequently asked questions
What exactly is Bloome?
Why put several agents in the same conversation?
Which models and coding agents can I use?
Can I build or clone my own agents?
What does it cost?
Are my conversations used to train AI models?
Where is my data hosted, and who else sees it?
Is there an API, and which devices are supported?
Is there a minimum age?
What about GDPR paperwork for a company deployment?
Should you pick Bloome?
Bloome makes one clear bet: the place where agents should work is the conversation, not a side panel. That bet pays off in a specific way — because several agents share the thread, one can be made to review another, and the second pass is what separates the product from a single-answer assistant. Around it sit the pieces you expect from a platform rather than a chat app: an agent builder, a public library you can clone from and publish to, installable skills, coding agents wired in through ACP, and native clients on five platforms.
The product is young, and it shows in useful ways as well as awkward ones. The first web archive of the domain dates from April 2026, the public launch was announced on 1 July 2026, the cloud agent is still in Beta, and the site itself already carries a fallback redirect from an earlier domain. Data commitments read better than average for a product this age: you keep ownership of your submissions, Bloome states it does not train models on your content, encryption and audit logging are described concretely, and the eight AI providers your messages may reach are named one by one.
What should give a professional buyer pause is everything the site does not say. There is no published price of any kind, only credits bought inside the app whose consumption rate the vendor may change. There is no postal address, no full legal name and no legal-notice page. There is no DPA, no Article 27 representative and no completed SOC 2 audit, while hosting sits primarily in the United States. For an individual or a small team wanting agents that argue with each other, Bloome is worth an evening. For a regulated organisation, the paperwork is not there yet.
- Choosing a selection results in a full page refresh.
- Opens in a new window.