convez.ai
MCP ServerAutomaçãoIntermediário
BrainGridAI

Radial MCP Server

BrainGridAI/radial

Rastreador de issues com API para agentes — controle seu backlog via IA

Ver no GitHub
InicialPluginScore · de 0 a 100

MCP server que integra o Radial, um rastreador de issues keyboard-first, com agentes de IA. Permite automação de fluxos de trabalho de planejamento, prototipagem e construção de features através de CLI, REST API e skills para Claude e outros agentes.

Instalação

$npx convez-ai BrainGridAI/radial

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
0
Forks
0
Acompanhando
0
Saúde
50%
Último commit
há 2 semanas
Versão
cli-v1.3.0
Issues abertas
0
Contribuidores
1

Casos de uso

  • Automação de planejamento e decomposição de features com agentes
  • Integração de fluxos de desenvolvimento com Claude Code e Codex
  • Scripting de operações em massa no rastreador de issues via terminal

Veredito

Solução robusta para times que querem automatizar gestão de issues sem vendor lock-in em IA. O design keyboard-first e a capacidade de trazer seu próprio agente via MCP são fortes diferenciais, especialmente para equipes técnicas que já usam Claude.

README

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

Radial

The fast, keyboard-first issue tracker with no AI bolted on.

No credits, no copilots, no surprises. Bring your own agent via a first-class CLI and MCP. $50 per user per year, flat.

npm  provenance  license

radial.build · Docs · Agent skills · Developers · MCP


This is Radial's developer surface: the radial CLI, the MCP server, the REST API, and the rd-* agent skills. Built to be scripted by you and your agents. The Radial app itself lives at radial.build; this repo is the home of the tooling, the skills and the releases.

Install

# npm (any platform with Node 18+)
npm install -g radial.build

# Homebrew (macOS / Linux)
brew install BrainGridAI/radial/radial

Then run radial.

Quickstart

radial auth                                              # device-flow sign-in (opens your browser)
radial create "Search flashes an empty state" -p high -l bug -a me
radial list --status open --assignee me
radial show RAD-219

Agent skills

Eleven skills that teach Claude Code and Codex how to work in Radial: plan a change into a parent and a few children, file the tree with its dependency edges, build it child by child leaving notes on each issue, and open a pull request that does not auto-close the siblings. They are prose, not scripts: no code runs on your machine that you did not read.

60 seconds:

radial skills install          # ~/.claude/skills and ~/.agents/skills
radial skills install --project  # .claude/skills and .agents/skills, for a repo
radial skills list

Then restart your agent and type /rd-setup (Claude Code) or $rd-setup (Codex).

Prefer another installer? Both of these read the same folders:

npx skills add BrainGridAI/radial
codex -- '$skill-installer --repo BrainGridAI/radial --path skills/rd-plan'

The workflow, in order

SkillWhat it doesExtension points
rd-setupCLI, sign-in, team, .radial/config.json, the CLAUDE.md / AGENTS.md block, MCP, the packextra-checks, extra-config, after-setup
rd-planAn idea becomes one parent and 2–4 children with acceptance criteria, written to .radial/plans/extra-questions, extra-sections, extra-checks, plan-bar, after-plan
rd-prototypeOne self-contained clickable HTML file, attached to the issue with a Preview URLextra-craft, extra-screens, save-bar, extra-notes, extra-checks
rd-issuesFiles the tree, wires blocked-by, reads every issue backextra-fields, file-bar, after-file
rd-buildWorks the tree child by child: in progress → build → gate → browse checks → notes → in reviewbefore-child, extra-checks, review-bar, after-child, after-tree
rd-verifyResolves and runs this repo's gate, then reads the summary line instead of the exit codegate-commands, extra-checks, green-bar
rd-prBranch, base sync, commit, pull request with sibling ids scrubbed, post-merge status repairextra-checks, extra-sections, merge-bar, after-pr

Hygiene, any time

SkillWhat it doesExtension points
rd-triageRanks open parents Now / Next / Later / Icebox from real signal; writes only on yesextra-signals, rank-bar, after-rank
rd-cleanupOpen children under done parents, decided on evidence: close or promote. Never deletesevidence-bar, extra-sweeps, after-sweep
rd-changelogCompleted issues become Feature / Improvement / Fix entries in .radial/changelog.mdextra-sources, entry-bar, output
rd-retroScores one finished build against the workflow, with evidence; proposes wrapper sectionsextra-rules, score-bar, after-retro

What lives where

PathWhat it isCommit it?
.radial/config.jsonThe team key for this repoyes
.radial/skills/<name>/Your wrappers: how your team extends a skillyes
.radial/changelog.mdThe changelog rd-changelog writesyes
.radial/plans/Plan bodies, scaffolding for rd-issuesno
.radial/prototypes/Prototype HTML before it is attachedno

Full reference, one page per skill, at radial.build/docs/skills.

Extending a skill

Yours does not replace ours. A wrapper adds rules to a built-in at named places, so you keep getting our updates and we never overwrite your conventions.

A wrapper is an ordinary skill folder with an extends field:

---
name: team-build
description: Our extra rules on top of rd-build, for the payments repo.
extends: rd-build
---

Read rd-build first, then apply the sections below at their hooks.

## before-child

- Refuse to start a child on a dirty working tree.

## review-bar

- A child may not move to in review until a screenshot of every browse check is
  attached to its issue.

Save it at .radial/skills/team-build/SKILL.md (this repo) or ~/.config/radial/skills/team-build/SKILL.md (you, everywhere), or let the CLI scaffold it with every heading already in place:

radial skills extend rd-build --name team-build

The rules, each with its reason:

RuleWhy
Sections are named places, not line numbersThe built-in can be rewritten without breaking your wrapper
A wrapper adds rules and raises bars; it cannot remove a step or lower a barOtherwise an extension could quietly disable the check that mattered
The nearest scope wins whole: project beats user, never mergedTwo half-applied rule sets is the worst of both
Two wrappers in one scope extending one skill: neither appliesSilent precedence is worse than a loud refusal
rd-* names are reservedSo an update never fights your wrapper for a name

radial skills install --project also copies this repo's wrappers into .claude/skills and .agents/skills, so /team-build is invocable directly.

Full guide: radial.build/docs/skills/extending.

Contributing a skill

The pack is MIT and lives in skills/. Copy skills/_template/SKILL.md, keep the section order, then:

npm run manifest   # regenerate skills/manifest.json
npm test           # the guards: zero dependencies, node --test

The guards enforce the things that rot: the manifest matches the folders, every hook marker is alone on its line, bar names are unique across the pack, no executable ships inside a skill folder, and this README lists every skill and every hook.

Commands

CommandWhat it does
radial authDevice-flow sign-in; stores a session under ~/.config/radial/
radial create <title>Create an issue. -p priority, -l label (repeatable), -a assignee, -t team
radial update <ID>Edit an issue. --add-label / --remove-label change one label; -l replaces the set
radial listList issues. --assignee, --status, --label, --team, --priority, --json
radial show RAD-219Full detail + comments
radial close RAD-219Close, with optional -m "message"
radial skills <sub>install · update · list · remove · extend for the rd-* agent skills
radial triageStep through the triage queue
radial search <query>The same fast index the app uses
radial branch RAD-219The suggested git branch name for an issue
radial attach <ID> <file>Attach a prototype (or file) and print its Preview URL
radial import --from linear|jira <file>Deep import (issues, labels, projects, parents, relations, comments). --dry-run first
radial exportFull workspace export. --format json|csv, -o file
radial mcpRun the MCP server over stdio for any MCP-capable agent

Add --json to any read command to pipe into jq and friends.

MCP

Point any MCP-capable agent (Claude, etc.) at Radial.

Remote (recommended) authorizes over OAuth in the browser, no keys to paste:

{ "mcpServers": { "radial": { "url": "https://mcp.radial.build" } } }

Local stdio runs the server yourself, using your radial auth session:

{ "mcpServers": { "radial": { "command": "radial", "args": ["mcp"] } } }

Tools: create_issue, update_issue, search_issues, list_issues, comment, close_issue, list_projects, triage_queue.

REST API

Base URL https://api.radial.build/v1. Authenticate with a Bearer API key (rk_…, created in Settings → Developers) or via OAuth 2.1.

curl https://api.radial.build/v1/issues \
  -H "Authorization: Bearer rk_full_xxxxxxxx"

Resources: issues, projects, search, triage, export, import, meta. Full reference at radial.build/developers.

Configuration

State lives in ~/.config/radial/config.json (mode 0600). Override endpoints with RADIAL_API_URL, RADIAL_AUTH_URL, RADIAL_MCP_URL, handy for self-hosting. The skills installer reads RADIAL_SKILLS_URL the same way.

License

MIT. See LICENSE.

Ver o original no GitHub

Bom para

Relacionados