feat(mcp): first-class MCP tab — catalog, GUI auth/probe/logs, per-tool gating

A Cursor-style MCP manager inside Capabilities, plus the backend it needs.

- Server list with brand/favicon avatars + live status dot and a capability
  summary (N tools, M prompts, K resources); Servers | Catalog views.
- Catalog: one-click install of Nous-approved servers with required-env prompts.
- GUI OAuth: Authenticate opens the system browser from the TTY-less backend and
  verifies a token actually lands; header/API-key servers are never pushed down
  OAuth; a dirty mcp.json can't drop a freshly-persisted auth field.
- Full-width mcp.json editor (ecosystem document format) + pinned stdio/agent
  LogTail; probes cached 5m and keyed by (profile, config) so revisiting never
  respawns the fleet or shows a stale probe.
- Whole-map persistence (PUT /api/mcp/servers) so deletes/toggles actually stick
  (the generic /api/config deep-merge could not remove keys).
- perf: MCP probe/auth no longer hold the global skills lock, so a slow stdio
  spawn can't stall every other request into a 15s timeout.
- per-tool include/exclude gating (lib/mcp-tool-filter) mirroring the CLI loader.
This commit is contained in:
Brooklyn Nicholson 2026-07-03 05:08:28 -05:00
parent 7e6d60aadc
commit 16aa09aca5
12 changed files with 1959 additions and 36 deletions

File diff suppressed because it is too large Load diff

View file

@ -73,9 +73,7 @@ describe('Hermes REST session helpers', () => {
})
it('uses a longer timeout for active profile refresh during desktop startup', async () => {
api
.mockResolvedValueOnce({ current: 'default' })
.mockResolvedValueOnce({ profiles: [] })
api.mockResolvedValueOnce({ current: 'default' }).mockResolvedValueOnce({ profiles: [] })
await refreshActiveProfile()

View file

@ -22,7 +22,6 @@ import type {
LogsResponse,
McpCatalogResponse,
McpServerSummary,
McpServerTestResponse,
MemoryProviderConfig,
MemoryProviderOAuthStatus,
MemoryStatusResponse,
@ -343,6 +342,7 @@ export function getLogs(params: {
file?: string
level?: string
lines?: number
search?: string
}): Promise<LogsResponse> {
const query = new URLSearchParams()
@ -362,6 +362,10 @@ export function getLogs(params: {
query.set('component', params.component)
}
if (params.search) {
query.set('search', params.search)
}
const suffix = query.toString()
return window.hermesDesktop.api<LogsResponse>({
@ -592,6 +596,49 @@ export function toggleSkill(name: string, enabled: boolean): Promise<{ ok: boole
})
}
export interface McpTestResult {
ok: boolean
error?: string
tools: { name: string; description: string }[]
/** Capability counts (absent on older backends / failed probes). */
prompts?: number
resources?: number
}
/** Connect to the server, list its tools, disconnect. Slow (spawns/handshakes
* for real) well past the 15s default fetch timeout. */
export function testMcpServer(name: string): Promise<McpTestResult> {
return window.hermesDesktop.api<McpTestResult>({
...profileScoped(),
path: `/api/mcp/servers/${encodeURIComponent(name)}/test`,
method: 'POST',
timeoutMs: 60_000
})
}
/** Replace the whole `mcp_servers` map (the mcp.json editor's save). Unlike
* `saveHermesConfig`, this REPLACES rather than deep-merges, so deletes,
* re-enables (dropping `enabled: false`), and removed nested fields persist. */
export function saveMcpServers(servers: Record<string, Record<string, unknown>>): Promise<{ ok: boolean }> {
return window.hermesDesktop.api<{ ok: boolean }>({
...profileScoped(),
path: '/api/mcp/servers',
method: 'PUT',
body: { servers }
})
}
/** Run the OAuth flow for an HTTP server opens the system browser and blocks
* until the user finishes (or gives up), hence the very generous timeout. */
export function authMcpServer(name: string): Promise<McpTestResult> {
return window.hermesDesktop.api<McpTestResult>({
...profileScoped(),
path: `/api/mcp/servers/${encodeURIComponent(name)}/auth`,
method: 'POST',
timeoutMs: 300_000
})
}
export function getToolsets(): Promise<ToolsetInfo[]> {
return window.hermesDesktop.api<ToolsetInfo[]>({
...profileScoped(),
@ -1033,16 +1080,6 @@ export function listMcpServers(): Promise<{ servers: McpServerSummary[] }> {
})
}
export function testMcpServer(name: string): Promise<McpServerTestResponse> {
return window.hermesDesktop.api<McpServerTestResponse>({
...profileScoped(),
path: `/api/mcp/servers/${encodeURIComponent(name)}/test`,
method: 'POST',
// Connect + list tools can be slow for stdio servers that boot a process.
timeoutMs: 60_000
})
}
export function setMcpServerEnabled(name: string, enabled: boolean): Promise<{ ok: boolean }> {
return window.hermesDesktop.api<{ ok: boolean }>({
...profileScoped(),

View file

@ -0,0 +1,74 @@
import { describe, expect, it } from 'vitest'
import { countEnabledTools, isToolEnabled, readToolsFilter, toggleToolInServer } from './mcp-tool-filter'
describe('readToolsFilter', () => {
it('returns empty when no tools object', () => {
expect(readToolsFilter({ command: 'x' })).toEqual({ exclude: undefined, include: undefined })
})
it('reads include/exclude and ignores non-string entries', () => {
expect(readToolsFilter({ tools: { exclude: ['c', 2], include: ['a', 'b', null] } })).toEqual({
exclude: ['c'],
include: ['a', 'b']
})
})
})
describe('isToolEnabled', () => {
it('enables everything with no filter', () => {
expect(isToolEnabled({ command: 'x' }, 'anything')).toBe(true)
})
it('include wins over exclude', () => {
const server = { tools: { exclude: ['a'], include: ['a'] } }
expect(isToolEnabled(server, 'a')).toBe(true)
expect(isToolEnabled(server, 'b')).toBe(false)
})
it('exclude disables listed tools', () => {
const server = { tools: { exclude: ['b'] } }
expect(isToolEnabled(server, 'a')).toBe(true)
expect(isToolEnabled(server, 'b')).toBe(false)
})
})
describe('toggleToolInServer', () => {
it('adds a fresh tool to a new exclude denylist when disabled', () => {
const next = toggleToolInServer({ command: 'x' }, 'a')
expect(next.tools).toEqual({ exclude: ['a'] })
})
it('re-enabling removes the tool and drops the empty exclude/tools', () => {
const next = toggleToolInServer({ command: 'x', tools: { exclude: ['a'] } }, 'a')
expect(next.tools).toBeUndefined()
})
it('respects include mode: toggling removes from include', () => {
const next = toggleToolInServer({ tools: { include: ['a', 'b'] } }, 'a')
expect(next.tools).toEqual({ include: ['b'] })
})
it('respects include mode: re-enabling adds back to include', () => {
const next = toggleToolInServer({ tools: { include: ['b'] } }, 'a')
expect(next.tools).toEqual({ include: ['b', 'a'] })
})
it('preserves sibling tools keys like resources/prompts', () => {
const next = toggleToolInServer({ tools: { resources: false } }, 'a')
expect(next.tools).toEqual({ exclude: ['a'], resources: false })
})
it('does not mutate the input server', () => {
const server = { tools: { exclude: ['a'] } }
toggleToolInServer(server, 'b')
expect(server.tools.exclude).toEqual(['a'])
})
})
describe('countEnabledTools', () => {
it('counts enabled tools out of a discovered list', () => {
const server = { tools: { exclude: ['b'] } }
expect(countEnabledTools(server, ['a', 'b', 'c'])).toBe(2)
})
})

View file

@ -0,0 +1,61 @@
// Per-tool MCP gating. A server's optional `tools.include` (whitelist) /
// `tools.exclude` (denylist) decide which discovered tools the agent registers
// — `include` wins, no filter means all. Mirrors `_register_server_tools` in
// `tools/mcp_tool.py`.
export interface McpToolsFilter {
exclude?: string[]
include?: string[]
}
type ServerConfig = Record<string, unknown>
const asNames = (value: unknown): string[] | undefined =>
Array.isArray(value) ? value.filter((v): v is string => typeof v === 'string') : undefined
const toolsObject = (server: ServerConfig | null | undefined): Record<string, unknown> => {
const tools = server?.tools
return tools && typeof tools === 'object' && !Array.isArray(tools) ? (tools as Record<string, unknown>) : {}
}
export function readToolsFilter(server: ServerConfig | null | undefined): McpToolsFilter {
const tools = toolsObject(server)
return { exclude: asNames(tools.exclude), include: asNames(tools.include) }
}
export function isToolEnabled(server: ServerConfig | null | undefined, name: string): boolean {
const { exclude, include } = readToolsFilter(server)
return include?.length ? include.includes(name) : !exclude?.includes(name)
}
// Toggle one tool, preserving the config's mode (include if present, else an
// exclude denylist). Empty lists — and an emptied `tools` — are dropped.
export function toggleToolInServer(server: ServerConfig, name: string): ServerConfig {
const { exclude, include } = readToolsFilter(server)
const key = include?.length ? 'include' : 'exclude'
const current = (key === 'include' ? include : exclude) ?? []
const names = current.includes(name) ? current.filter(n => n !== name) : [...current, name]
const tools = { ...toolsObject(server) }
if (names.length) {
tools[key] = names
} else {
delete tools[key]
}
const next = { ...server }
if (Object.keys(tools).length) {
next.tools = tools
} else {
delete next.tools
}
return next
}
export const countEnabledTools = (server: ServerConfig | null | undefined, names: string[]): number =>
names.filter(name => isToolEnabled(server, name)).length

View file

@ -509,9 +509,17 @@ export interface AnalyticsResponse {
summary: AnalyticsSkillsSummary
top_skills: AnalyticsSkillEntry[]
}
/** Per-tool-name call counts. Absent on older backends. */
tools?: AnalyticsToolEntry[]
totals: AnalyticsTotals
}
export interface AnalyticsToolEntry {
count: number
percentage: number
tool: string
}
export interface AnalyticsSkillEntry {
last_used_at: null | number
manage_count: number
@ -640,6 +648,10 @@ export interface SkillInfo {
description: string
enabled: boolean
name: string
/** Total observed activity (use + view + patch). Absent on older backends. */
usage?: number
/** 'agent' = learned/local (editable), 'bundled' = ships with Hermes, 'hub' = installed. */
provenance?: 'agent' | 'bundled' | 'hub'
}
export interface ToolsetInfo {