Skip to content
View 2nd1st's full-sized avatar

Highlights

  • Pro

Block or report 2nd1st

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
2nd1st/README.md
2nd1st — The second author was here first

2nd1st isn't a person. It's what's left when a person and a set of agents have worked together long enough that separating the two stopped being interesting.

Everything below came out of that arrangement — tools that were missing, built by whoever was closer to the problem.


Featured

Persistent, interactive MCP apps for any MCP host.

The AI builds an app once — and it's still there tomorrow. Data outlives the conversation, so nobody re-pastes a todo list into a fresh chat. Ships an engine, a 22-app App Store, and a hosted instance at openmcp.app.

MCP  SQLite  TypeScript  persistent UI

Brings the above into DeepSeek Harness — apps become sidebar containers with their own sessions, plus an agent status strip and inline widget rendering.

DSH  MCP apps  JavaScript


LLM infrastructure

api-log  ·  stars

Transparent trace recorder for LLM gateways. Sits in front of sub2api / CLIProxyAPI / new-api and records every request and response without touching the client. Written in Go.

Go  observability  reverse proxy

The reading end of it. A Svelte 5 SPA that makes those traces navigable — and it plugs into any compatible JSONL or SQLite store, not only the one it ships with.

Svelte 5  trace viewer  SPA


Agent capability

A vocabulary of visual forms, written for AI to consume. 406 catalogued components with trimmed schemas and boot snippets, so a model writes thin assembly code instead of reimplementing a Sankey from scratch. Explicitly tells the AI when not to reach for it.

TypeScript  catalog  AI-native docs

One identical task, every frontier model. Deterministic scoring across models × coding harnesses × effort levels, graded on visual and structural similarity — so the comparison survives scrutiny. Live results →

benchmark  evaluation  Next.js

A five-brain cognitive architecture for agents meant to run long, not just answer once.

TypeScript  agent architecture


Working with

Agents     MCP · Claude Code · DeepSeek Harness · multi-agent orchestration
Backend    Go · Python · SQLite · gateway internals & reverse proxies
Frontend   TypeScript · Svelte 5 · Next.js
Infra      Incus · self-hosted everything

secondfirst  ·  openmcp.app  ·  admin@secondfirst.ai

Pinned Loading

  1. api-log api-log Public

    Transparent LLM gateway trace recorder · sub2api / CLIProxyAPI / new-api 的 LLM 网关流量录制器

    Go 11 2

  2. weather-card-benchmark weather-card-benchmark Public

    One identical task, every frontier model — how differently do they build it? A deterministic weather-card benchmark across models × coding harnesses × effort levels, scored on visual/structural sim…

    HTML 2

  3. open-mcp-apps open-mcp-apps Public

    Persistent, interactive MCP Apps for any MCP host: the AI builds an app once, you both reuse it — data outlives the conversation. Ships an engine, a 22-app App Store, and a hosted version at openmc…

    JavaScript 30

  4. open-visualize open-visualize Public

    TypeScript 1