convez.ai
MCP ServerProdutividadeIntermediário
zcag

tela — Wiki em Markdown com MCP Server Nativo

zcag/tela

Wiki markdown colaborativa com MCP server para IA como cidadão de primeira classe

Ver no GitHub
InicialPluginScore · de 0 a 100

Wiki colaborativa self-hostable construída em Markdown com suporte nativo a agentes IA via MCP server. Oferece busca semântica + full-text, edição em tempo real com Yjs, controle de versão e compartilhamento público. Stack moderno (Go + PostgreSQL + React) pronto para deploy em Docker Compose.

Instalação

$npx convez-ai zcag/tela

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
98
Forks
13
Acompanhando
1
Saúde
42%
Último commit
há 6 dias
Versão
v0.11.3
Issues abertas
0
Contribuidores
1

Casos de uso

  • Bases de conhecimento da equipe com leitura/escrita automática por agentes IA
  • Documentação técnica colaborativa com histórico e busca semântica
  • Wiki corporativa self-hosted com segurança total sobre dados

Veredito

Solução robusta e bem arquitetada que quebra a barreira entre humanos e agentes IA em wikis colaborativas. Excelente para times que querem documentação inteligente sem perder controle dos dados.

README

A tradução para o português deste README ainda está na fila.

tela

A self-hostable, markdown-native team wiki — spaces, nested pages, a Milkdown editor, ranked full-text + semantic search, live collaboration, comments, page history, public sharing, and a first-class MCP server so AI agents are first-class citizens alongside humans.

Public instance: https://tela.cagdas.io

Stack

PartTechLocation
BackendGo + PostgreSQL (pgx/v5 stdlib), database/sql (no ORM), embedded migrationsbackend/
FrontendReact 19 + TypeScript + Vite + Tailwind v4 + Radix + Milkdown + TanStack Query/Router + Oramafrontend/
CollabYjs + y-prosemirror over a custom WebSocket transport (not y-websocket)frontend/
MCP serverTypeScript (@modelcontextprotocol/sdk), published to npm as tela-mcpmcp/
DeployDocker Compose — backend, frontend (nginx), Postgres, Caddy proxy; data on a Postgres volumedeploy/

pages.body is canonical markdown forever; there is no block table. The editor is Milkdown; Yjs is an overlay that rebases onto the markdown on save.

Quickstart (local dev)

make dev        # backend (go run ./cmd/tela, :8080) + frontend (vite, :5173) in parallel

The Vite dev server proxies /api → backend :8080. make dev boots a local Postgres (tela-dev-pg on :55433) and points TELA_DATABASE_URL at it; the schema is migrated automatically on backend start (embedded migrations run by db.Migrate() — no separate migrate step). Open http://localhost:5173.

Run parts individually:

make be-dev     # backend only (go run ./cmd/tela)
make fe-dev     # frontend only (vite)
make storybook  # Storybook — the component dev surface (:6006)

Production (Docker Compose)

make setup                           # write deploy/.env from the example with generated secrets
$EDITOR deploy/.env                  # set admin creds, public base URL, SMTP
make up                              # build + start the stack (Docker only — no host Node)
make logs                            # tail logs
make down                            # stop

The stack publishes a single host port 8780 (Caddy). make setup fills the load-bearing secrets (TELA_SHARE_SECRET, TELA_API_KEY_SECRET, TELA_PG_PASSWORD); set TELA_PUBLIC_BASE_URL and the TELA_ADMIN_* / TELA_SMTP_* values yourself. If you leave the HMAC secrets unset entirely, tela now generates and persists them on first boot (stable across restarts).

Full setup, TLS, backups, and upgrades: see docs/self-hosting.md and the operations runbook docs/operations.md. deploy/.env.example documents every variable.

Make targets

make setup      # first run: write deploy/.env from the example with generated secrets
make dev        # backend + frontend in dev mode (parallel, no compose)
make be-dev     # backend only           make fe-dev     # frontend only
make storybook  # Storybook
make up         # build + start the compose stack on :8780 (stamps version/commit)
make down       # stop          make logs   # tail logs       make build  # rebuild images
make backup     # dump Postgres to ./backups   make restore FILE=...  # restore a dump
make clean FORCE=1            # stop AND delete volumes (destroys data)
make test-mcp-integration     # boot stack + run MCP↔backend E2E + tear down
make help

Note: there is no make lint or make migrate — migrations run automatically on backend start. make test runs the backend suite (below).

Tests

make test                       # backend (boots a throwaway Postgres; per-test isolated DBs)
cd mcp && npm test              # MCP unit (mocked fetch) + npm run test:smoke / test:integration

The frontend has no unit-test harness today (no jsdom / vitest config). CI (.github/workflows/ci.yml) runs the MCP↔backend integration suite on changes to backend/, mcp/, deploy/, or the Makefile.

Documentation

  • docs/architecture.md — system shape, repo layout, data model, and per-subsystem load-bearing details
  • docs/decisions.md — why PostgreSQL, custom collab transport, MCP-as-thin-client, etc.
  • docs/api.md — REST API surface
  • CLAUDE.md — conventions, hard rules, and the gotchas — read before contributing
  • mcp/README.md — the MCP server (install, tool catalog, troubleshooting)

History

tela was originally built by an autonomous agent system ("forge") and is now under conventional development. The old forge workspace (plan, run logs, task/idea history, research output under docs/output/) is archived out-of-tree — reference only, not part of this repo. This repo's code, docs/, and git history are the source of truth.

License

Copyright © tela contributors. Licensed under the GNU Affero General Public License v3.0 (AGPL-3.0). You may self-host, modify, and redistribute tela under its terms — in particular, if you run a modified version as a network service you must offer your users the corresponding source. For a commercial license without AGPL obligations (e.g. to embed or offer tela as a closed service), contact the maintainer.

"tela", the tela name, and the tela logo are trademarks and are not licensed under the AGPL — see TRADEMARK.md. You may run and fork the code, but you may not use the tela branding for a redistributed or hosted version without permission.

Ver o original no GitHub

Bom para

Relacionados