Frequently asked questions
Everything about how Tseha works - agents, setup, roles, security, and plans.
Looking for the side-by-side plan comparison? See pricing.
General
- What is Tseha?
- Tseha is a governed development-context platform. You define your engineering standards, approved components, design tokens, and architectural rules once, and Tseha exposes them to AI coding agents over the Model Context Protocol (MCP) - so agents follow your team's conventions on every request, across every project.
- What is MCP, and why does it matter?
- Model Context Protocol (MCP) is an open standard from Anthropic that lets AI coding agents query external data before they act. Instead of stuffing a long static prompt into every repository, your agent fetches only the standards relevant to the current task - always current, centrally managed.
- How is this different from a CLAUDE.md file?
- A CLAUDE.md lives in one repository and has to be copied and kept in sync by hand across every project. Tseha is centralized and queryable by topic - every repository that adds a single .mcp.json gets the same up-to-date standards automatically, with no copy-paste and no drift.
AI agents & setup
- Which AI coding agents are supported?
- Tseha is client-agnostic: it is a standard MCP server, so any MCP-compatible agent works - Claude Code, Cursor, GitHub Copilot (with MCP support), and any other tool that implements the protocol. There is no per-agent integration to install; agents discover the server from your .mcp.json.
- How do I connect my repository?
- Add a single .mcp.json file pointing at the Tseha MCP server. Your agent authenticates with your account on first use, then queries Tseha automatically on the next task it performs.
- How long does setup take?
- About two minutes: sign in with Google, create an organization, define your first standard, and add .mcp.json to your repository. No credit card required to start.
- Can CI pipelines and automation tools use Tseha?
- Yes, on the Team and Enterprise plans. Interactive agents authenticate as the signed-in member, but non-interactive consumers - CI jobs (GitHub Actions, GitLab CI, Azure Pipelines), AI code-review bots, and automation platforms such as n8n, Make, or Zapier - have no browser to sign in with. For them, an organization admin creates a long-lived, read-only API token and passes it as a Bearer header. Tokens can be scoped to all projects or a specific set, given an optional expiry, are shown only once, and can be revoked instantly.
Roles & permissions
Every organization has four roles. Read access and write access over MCP are both governed by the role, and project visibility is enforced on top of it.
| Role | What it can do |
|---|---|
| OwnerThe organization’s super-admin - exactly one. |
|
| AdminRuns the organization day to day. |
|
| DeveloperBuilds with the standards and contributes to them. |
|
| UserRead-only consumer of the design system. |
|
- Who can read and who can write over MCP?
- Owners, Admins, and Developers have write access; Users are read-only. Membership is re-checked on every MCP call, so a role change or a revoked member takes effect immediately - no token rotation needed.
- How is project access controlled?
- Owners and Admins can access every project in the organization. Developers and Users only see the projects they have been explicitly assigned to. This is why an agent only ever lists the projects the signed-in member is entitled to.
- Can I change someone’s role later?
- Yes. Admins and Owners can change a member’s role or remove them at any time from the admin panel. There is always exactly one Owner; ownership can be transferred to another member.
Security & data
- Does my source code ever go to Tseha?
- No. The MCP server is read-only and outbound: it serves your standards, components, and tokens to your agent. It has no tool that reads, receives, or stores your code, and your code never flows to Tseha. The only thing your agent sends is a short search description when it looks up a component by use case.
- Where is my data stored?
- On infrastructure operated by Tseha, in the EU. All content is encrypted at rest and in transit, and integration tokens (for example Figma) are encrypted with authenticated encryption (AES-256-GCM), each record using a unique initialization vector.
- How does authentication work?
- You sign in with Google. Interactive agents connect over OAuth with PKCE, and the issued token is scoped to your organization and role. For non-interactive (machine) access, admins can issue read-only, project-scoped access tokens. Administrative token actions are recorded in an audit log.
Plans & pricing
See the full side-by-side comparison on the pricing page. The essentials:
- Is there a free plan?
- Yes. The Starter plan is free forever - 1 project, up to 3 members, up to 3 packages, up to 5 coding standards, and a full MCP server included. No credit card required.
- How much do paid plans cost?
- Team is €39 per month or €32 per month when billed yearly (about 18% off). Enterprise is custom-priced - contact us for a quote. You can upgrade, downgrade, or cancel at any time.
- What do paid plans add?
- Team and Enterprise unlock the AI-backed features - AI package import (npm & ZIP), semantic component search, AI-generated standard summaries, Figma token sync, update & breaking-change tracking, and the audit log - plus long-lived API tokens for CI and automation, along with higher limits. Enterprise adds unlimited scale, custom MCP rate limits, and dedicated support.
- Why are those features gated?
- The paywall sits on real marginal cost: semantic search, npm/ZIP analysis, and AI summaries are model-backed. The core value - your standards, served over MCP - is available on every plan, including the free one.
Still have a question? Ask us directly.