Give your agent Capabilities, not a tool wall

Turn sprawling MCP servers into focused capability cards. Your agent can start with one typed route, zoom into tools only when needed, and keep huge schemas out of the prompt until they matter.

Tool wall: 215 tools
Caplets: 7 cards
Code Mode: 72% fewer tokens

Install & run setup GitHub repo npm package

Quick start More setup options

01
Install Caplets npm install -g caplets
02
Wire up your agent caplets setup

Setup

Run setup, then add one Caplet.

caplets setup lets you choose which agent integrations to wire up. Start with the no-auth OSV Caplet, then add larger surfaces like GitHub once the flow is working.

Explore more Caplets

Add OSV

A small explicit HTTP API for vulnerability lookups by package, purl, commit, or batch query.

Start here: OSV is public, no-auth, and shows the capability surface without OAuth friction.

osvinspectsearch_toolsget_toolcall_tool

OSV is public and does not need auth; check Node 24+ and caplets setup.

caplets setup
caplets install spiritledsoftware/caplets osv
codex "try using the osv caplet"

Add GitHub

A huge hosted MCP surface for repositories, issues, pull requests, branches, commits, and reviews.

Use it when the value is avoiding a giant GitHub tool wall.

githubinspectsearch_toolsget_toolcall_tool

If setup fails, check Node 24+, caplets setup, and GH_TOKEN.

export GH_TOKEN=github_pat_...
caplets setup
caplets install spiritledsoftware/caplets github
codex "try using the github caplet"

Add Sourcegraph

Hosted code search for finding examples, references, and implementation patterns across repositories.

Use it when the agent should search code first, then inspect only the matching operations.

sourcegraphinspectsearch_toolsget_toolcall_tool

If setup fails, check Node 24+, caplets setup, and Sourcegraph OAuth login.

caplets setup
caplets install spiritledsoftware/caplets sourcegraph
caplets auth login sourcegraph
codex "try using the sourcegraph caplet"

Benchmark proof

Real-world large MCP stack: GitHub, Context7, DeepWiki, Git, filesystem, Playwright, ast-grep, language-server, and web-search MCP servers.

Run June 2026 with the real-world large MCP suite, openai-codex/gpt-5.5, 10 tasks, 2 runs per task, and a large no-fixture MCP stack.

passed: 10/10
avg tokens: 236,803
vs vanilla: 72.0% fewer

Mode	Passed	Total Tokens	Tool Surface Tokens
Caplets Code Mode	10/10	236,803	31,166
Caplets progressive + Code Mode	10/10	422,861	124,772
Caplets progressive	10/10	461,171	103,776
Executor.sh	10/10	675,842	24,570
Direct vanilla MCP	10/10	846,048	457,818

direct flat tools 215

Caplets cards 7

payload reduction 79.9%

Caplets Code Mode used 72.0% fewer request + output tokens than Direct vanilla MCP and 65.0% fewer than Executor.sh. Caplets progressive also used 45.5% fewer than Direct vanilla MCP and 31.8% fewer than Executor.sh, while all listed modes passed 10/10 tasks.

Read benchmark method

Remote Caplets server

Auth into tools once. Use them from every agent.

Run Caplets as a small HTTP service. Provider tokens and OAuth state stay with that server; Codex, OpenCode, Pi, Claude Code, and any MCP client connect to the same capability surface.

Without remote Caplets

Each agent client needs its own provider tokens, OAuth login, and local MCP wiring.

Repeat auth per client
Duplicate secrets
Debug setup in every agent

With remote Caplets

One server holds provider auth; every agent connects to the same capability surface.

Auth once on the server
Reuse from Codex, OpenCode, Pi, Claude
Inspect, search, schema, call everywhere

Server owns auth GitHub, Sourcegraph, OSV

Caplets exposes inspect · search · schema · call

Agents reuse it Codex · OpenCode · Pi · Claude

Serve once

CAPLETS_SERVER_URL=https://caplets.example.com/caplets
CAPLETS_SERVER_PASSWORD=...
caplets serve --transport http

Use remotely

CAPLETS_MODE=remote
CAPLETS_SERVER_URL=https://caplets.example.com/caplets
CAPLETS_SERVER_PASSWORD=...
opencode

MCP: /caplets/mcp
Control: /caplets/control
Health: /caplets/healthz