Tseha.io MCP server icon

Tseha.io

by Tseha.io

HIPAA CompliantSOC2 ReadyISO 27001 Ready
Productivity9 tools

Give an AI coding agent your organization's real component library, design tokens and engineering standards instead of letting it guess. 9 tools, all read-only and annotated as such by the server, OAuth sign-in, no application scopes.

Verified connector

Listed by Anthropic as a partner connector in its Connectors Directory.

Connection checked by Agentman on .

Anthropic states this reflects the level of review a connector received, not a security audit.

Connect Tseha.io via MCP

https://tseha.io/mcp

Works in any MCP-compatible client. In Agentman Studio it is one click — no config file to edit.

Tseha.io Tools & Capabilities (9)

list_projectsRead-only

Lists the projects in this organization, each with its id, name, and framework. Read-only, takes no arguments, and returns only the projects the calling credential may see. Call it first in a session: the returned id is the project_id that every other tool in this server requires, and guessing that id fails. It does not return components, packages, or tokens - use list_components, list_packages, or get_style_tokens for those.

list_packagesRead-only

Lists the component packages assigned to a project, each with its name, version, and description, and marks which one is active. Read-only. Use it to decide which npm package to import from, or to read the installed version before calling get_component_updates. It lists packages, not their contents - use list_components to enumerate components and get_component for one component API. project_id is the only argument and it does more than filter output: assignment and active-package status are per project, so one organization package can be listed as active here and absent from a sibling project.

list_componentsRead-only

Lists every UI component in the project's design system, each with its name, source package, import path, and a short note on its intended use. Read-only and unpaginated: it returns the whole catalog across all packages assigned to the project, so expect a large response on big design systems. Use it to see what exists before building UI, and to get the exact component name other tools need. Do not use it to find a component by need - search_components ranks by description - and do not use it for props, examples, or Figma data, which only get_component returns. project_id is not a display filter: it selects the assigned packages, so two projects in one organization legitimately return different catalogs.

get_componentRead-only

Returns the full specification for one component: import path, props schema, usage example, anti-patterns, when-to-use and when-not-to-use notes, Figma spec, and dependencies. Read-only. Call it before writing code that uses a component, so props and import path come from the design system instead of memory. The name argument must be the exact component name as returned by list_components or search_components (case-sensitive, no package prefix); an unknown name returns a not-found error rather than a near match, so resolve the name first with search_components if you are guessing. The result may carry a provenance warning when the record was LLM-generated and not yet reviewed - treat those fields as unverified. It covers one component at a time; use list_components to enumerate and get_component_updates for version-to-version changes.

search_componentsRead-only

Searches the components of every package assigned to the project by use case, behavior, or description, and returns up to 10 ranked matches with a relevance score. Read-only. Use it when you can describe what the UI must do but do not know the component name. The query is free-text describing the need, such as "date range picker" or "dismissible inline warning" - not a component name, not a glob, and not a props expression. Do not use it when the exact name is known (get_component is direct and returns the full spec) or when you want the complete catalog (list_components). Semantic ranking needs AI features on the Team plan or above; on other plans it falls back to name and import-path matching and the score is null. No close match returns an empty result list, not an error.

get_styleRead-only

Returns metadata for the project's active style: the style name, its Figma file key, when it was last synced, and a token count per available category (such as color, typography, spacing). Read-only and small - it names categories but returns no token values. Use it to inspect the project theme, or to learn which categories are populated before requesting values; call get_style_tokens for the values themselves. project_id is the only argument and it is not a plain filter: the active style is resolved per project, where a project-level override beats the organization default, so sibling projects can report different style names and different populated categories. When the project has no style configured at either level it returns a null style with a note rather than an error.

get_style_tokensRead-only

Returns the project's design tokens with their names, values, and token paths. Read-only. Use it to source exact visual values for styling instead of inventing hex codes, spacing, or radii. The category argument takes exactly one of: color, typography, spacing, radius, shadow, blur, all; omitting it behaves the same as "all". Pass a single category when you know which one you need - "all" returns the largest response this server produces. Use get_style first if you only need to know which categories exist, and get_standards for coding conventions, which are not tokens. A project with no style configured returns an empty token list with a note, not an error.

get_standardsRead-only

Returns the organization's development standards: coding conventions, project structure, and framework-specific rules. Read-only. Call it before writing or reviewing code, so the result follows this organization's rules rather than general defaults. Call it first without a section to get an index of available sections, each with a note on what it covers, then call again with one section id copied from that index; inventing a section id returns a not-found error naming that step. Request only the sections a task needs - the full content of one section can be long. The framework argument is deprecated: use section with the "framework:" prefix instead. It returns prose rules, not data - use get_style_tokens for visual values and get_component for component APIs.

get_component_updatesRead-only

Returns per-component breaking changes and a migration guide between an installed version of the project's active component package and the latest published version. Read-only: it reports an upgrade, it never performs one. Use it when planning or reviewing a component-library upgrade, or before bumping the package version in a manifest. current_version is the exact semver currently installed, e.g. "1.0.0" - not a range, so no leading "v", "^", or "~"; read it from list_packages when unsure. Passing the latest version returns an empty change set. It is not a per-component changelog and not a current spec - use get_component for one component today, and list_packages for what is installed. Requires update tracking, a Team plan feature; on lower plans it returns an upgrade-required error instead of data.

Read from the server on 2026-08-23, including each tool's own safety annotations.

Limits

  • No application scopes exist. Neither descriptor declares scopes_supported, so the consent screen offers no granularity. The practical cost is low here — every tool is read-only — but a token grants the full read surface for every project the member can reach, and nothing narrower can be issued interactively.
  • Two tools are plan-gated. get_component_updates needs update tracking on Team or above and returns an upgrade-required error otherwise. search_components needs AI features for semantic ranking; below that it degrades to name and import-path matching with score: null rather than failing, which is the quieter failure of the two.
  • Free-plan ceilings are low. Starter allows 1 project, 3 members, 3 packages and 5 standards capped at 2,000 characters each. Team raises that to 10 projects, 15 members, 10 packages and 20 standards at 6,000 characters.
  • Responses truncate. A single tool response is capped at roughly 25,000 tokens. Tseha documents list_components as unpaginated across all assigned packages, so a large design system is exactly the case where it truncates — the documented remedy is calling get_component on one component instead.
  • Rate limit is 120 requests per minute per token, sliding window, returning HTTP 429 with Retry-After and JSON-RPC error -32002.
  • The component data may be LLM-generated. get_component can return a provenance warning marking a record as LLM-generated and not yet reviewed. An agent that reads props from Tseha and writes code against them is trusting whatever your organization published or generated, and only the warning distinguishes the two.
  • Machine tokens need a paid plan. CI and automation access requires long-lived API tokens, which Tseha gates to Team and Enterprise. There is no non-interactive path on Starter.
  • A deprecated parameter is still accepted. get_standards still takes framework, which its own schema marks deprecated in favour of section. Agents that learned the old shape will keep using it silently.
  • We could not observe tool behaviour, only declarations. No tool was called — the hard rule against calling tools on a third-party server applies. Everything above about what a tool returns comes from the server's own schemas and Tseha's documentation, both of which are claims.

Frequently asked questions

No. All nine tools carry the server's own readOnlyHint annotation set to true, and Tseha's documentation states the MCP server is read-only for every role, including Owner. Editing standards, components and tokens happens in the web admin panel instead. There is no write tool to gate.

No. Tseha's FAQ states the server has no tool that reads, receives or stores your code, and no tool schema accepts code as a parameter. The only content your agent sends is a short free-text description when search_components looks up a component by use case, plus a project UUID.

None that you can choose between. Neither the RFC 9728 resource descriptor nor the authorization server metadata declares scopes_supported at all, checked live on 2026-08-23. Access is bounded by your organization membership and role instead, which Tseha re-checks on every single request.

By design, so a client can confirm compatibility before sign-in. An anonymous handshake on 2026-08-23 returned all nine tool definitions with full schemas, while prompts/list and resources/list returned error -32001 Unauthorized. Calling any tool without a bearer token is refused.

Two of the nine tools degrade or stop on the free Starter plan. get_component_updates requires update tracking, a Team feature, and returns an upgrade-required error below it. search_components still answers but falls back to name and import-path matching with a null relevance score when AI features are off.

Centralisation and querying by topic. Tseha's FAQ states a CLAUDE.md lives in one repository and must be copied and kept in sync by hand, whereas every repository adding one .mcp.json file gets the same current standards. The agent also fetches only the section a task needs rather than one static block.

120 requests per minute per token, in a sliding window, per Tseha's documentation. Exceeding it returns HTTP 429 with a Retry-After header, surfaced over JSON-RPC as error -32002. Enterprise plans can configure custom limits. A single tool response is also capped at roughly 25,000 tokens and truncated beyond that.

Yes, on Team and Enterprise plans. Tseha states that non-interactive consumers such as GitHub Actions or code-review bots have no browser to sign in with, so an admin issues a long-lived read-only API token passed as a Bearer header. Tokens can be scoped to specific projects, given an expiry and revoked instantly.

Sources

Use in Agentman

Connect once and your agents call these tools on their own — on a schedule, in a workflow, with nobody at the keyboard.

Open in Agentman Studio

Server Info

Category
Productivity
Developer
Tseha.io
Tools
9
Domain
tseha.io

Using Claude Desktop or another MCP client? Setup docs — the connection URL above works anywhere.