Built first for a developer-driven workflow — with a clear team layer on top
SpecViber's shared working layer is already strong for individual developer-and-agent work, and the same model extends directly to teams. As team support rolls in, the value isn't just that more people get access — the development work itself becomes easier to understand, coordinate, and steer together.
Team value isn't just about more users. It's about shared readability, fewer collisions between parallel AI work, and clearer forward control.
When more people share the same specs, tasks, status, and relations, the project becomes easier to understand, coordinate, and steer together. Coherence first. Collaboration on top of coherence.
Shared metadataLocal contentOne active task per agentReview-gate before close
How team mode works
Shared metadata. Local content.
Team members share a common server for metadata — tasks, specs-index, relations, status, assignments — while each developer keeps their markdown files and MCP server running locally. The project still lives in each developer's repo. The team just shares the operational layer on top of it.
Workspace membership and roles (admin / editor / viewer)
What stays local
Each developer's markdown files and repo
Each developer's MCP server and local agent
Private drafts and in-flight edits until they're registered
Each developer's IDE session and working context
The team layer
What the team layer actually adds
Shared project view
The same specs, tasks, status, and progress — for everyone on the team.
Clearer ownership
It becomes obvious who is doing what, what's next, and what's blocking.
Comments in the right context
Discussions stay attached to the correct document, task, or action spec.
Smoother handoffs
Work can move between human and AI — and between teammates — without losing context.
Better forward control
Easier to follow development streams and keep momentum without creating chaos.
Why team vibe-coding doesn't collapse
Concrete mechanisms, not just promises
Parallel AI work in a team can collide fast: two agents editing the same files, missing review, untraceable changes. SpecViber has explicit coordination rules built into the workflow, not just principles on a slide.
One active task per agent (hard lock)
No agent — human or AI — can hold more than one in_progress task at a time. This eliminates the most common source of parallel collisions and forces work to be broken down into clear sequential steps.
Overlap detection on file scope
Tasks declare their file scope. The system warns — or blocks — when two active tasks touch overlapping files, so merge conflicts don't form silently in the background.
Branch-per-task automation
A task picked up for work lives on its own branch until it closes. Automatic checkpointing before work begins means any state is recoverable, even if an agent takes a wrong turn.
Review-gate before close
Tasks with unresolved checkboxes or @user items can't be closed as done — they go to waiting. AI work doesn't bypass human review on anything marked for human check.
Spec-driven context at pickup
When an agent picks up a task, it receives the full context in one call: the task, its source spec, related documents, prior notes, and unresolved comments. Context isn't reconstructed from a chat log every session.
Activity log per user and agent
Every status change, comment, and document edit is attributed and timestamped. A team can trace how a piece of work actually moved through — and who (or what) touched it.
How teams run SpecViber
Self-hosted or hosted
Teams choose where the shared metadata server lives. The architecture and code are the same either way.
Self-hosted team
The team runs its own SpecViber server — on an internal VPS, a company server, or any PHP + PostgreSQL host. You keep full control over data location, auth, and upgrade cadence. Same codebase as the hosted version.
Good fit when data locality, procurement, or internal compliance make self-hosting the simpler path.
Hosted team
SpecViber hosts the team's metadata server. The team admin invites members, members connect their local MCP and files. No VPS to operate. The team is billed on a subscription.
Good fit when you'd rather focus on the work than on running servers.
Honest status
Where team mode is today
The team architecture is decided and the codebase is shared with the single-developer path. The team MVP is being built — we describe it here so fit can be evaluated honestly.
Core architecture (shared metadata, local content)Decided
Coordination mechanisms (hard lock, overlap, branch-per-task, review-gate)Available in single-developer mode
Organisation / team / member model with rolesDesigned, in delivery
Self-hosted team setup guidePlanned
Hosted team tier with billingPlanned
If you're evaluating SpecViber for a team today, talk to us. Early team setups help shape the MVP and get priority support.
Why this is "the next step" and not the homepage opening
Team value is strongest when the foundation is already clear. Documents, tasks, status, relations, and verification must first function as a shared reality before you scale them across people. Selling the team story first would undersell that foundation.
Coherence first. Then collaboration on top of coherence.
FAQ
Why speak first to a developer and not to a whole team?
That's the strongest and most credible public fit right now. The team layer becomes better when the single-developer foundation is already clear — and most team wins compound from a good single-developer experience.
Is team mode available today?
The architecture is decided and much of the codebase is shared with the single-developer path. The organisation/team/member model and role-based access are in delivery. If you want to be one of the first teams on it, contact us — early setups help shape the MVP.
What does "shared metadata, local content" mean?
Team members share one server for tasks, specs-index, relations, status, and assignments. Each member keeps their own markdown files and runs their own MCP server locally. Your repo stays where it is — the team just shares the operational layer on top.
Self-hosted or hosted — how do we choose?
Self-hosted gives you full control over data location, auth, and upgrade cadence — suited to teams with compliance or procurement needs. Hosted means we run the server; you focus on the work. Same code, same features. Contact us to discuss which fits.
Won't multiple AI agents collide with each other?
That's exactly what the coordination mechanisms prevent. One active task per agent is a hard rule. File scope is declared per task, and overlaps are detected. Work lives on its own branch until closed. Review-gates block uncontrolled closure. These aren't guidelines — they're built in.
What's the real team win?
Not just shared access, but shared readability, clearer ownership, comments in the right context, fewer collisions between parallel AI work, and better forward control of the whole development stream.
Want to talk about team fit?
Tell us how your team works today, where coordination breaks down, and whether self-hosted or hosted would fit better. Early team setups get priority support as the MVP rolls in.
