Pier
A gateway between you and your codebase. Pier is an AI coding agent that lives in your terminal: it reads your repo, plans across files, edits code, runs commands, and ships.
Live platform
What it is
Pier is a model-agnostic coding agent for the terminal, priced from $5/₹499 a month, with bring-your-own-key use free. Behind a single command-line binary sits a serious agentic harness: multi-file planning, streaming edits, sandboxed command execution, project memory, and resumable sessions that sync across devices. Privacy is a feature, not a footnote: no analytics, no crash reporting, and keys that never leave the machine.
Building it meant shipping three production systems at once: an interactive terminal UI with an agentic tool loop, a commercial web platform with subscriptions and credit metering, and a high-throughput API bridge that makes a dozen-plus models from four providers look like one. We designed, built, and launched all of it in-house.
The three systems
The CLI
A terminal coding agent in TypeScript on Bun, compiled to a single self-contained binary for macOS, Linux, and Windows. React + Ink TUI, a native tool suite (file edits, search, shell, subagents, skills), MCP client, and OS-level sandboxing.
The Web Platform
A Next.js App Router site carrying the full account surface: auth, device-code pairing that logs the CLI in from the browser, subscriptions, credit dashboards, and dual-currency billing across USD and INR.
The Bridge Server
A Go service that translates the OpenAI Responses API to Chat Completions and back, so the CLI sees one uniform provider while turns route across Sarvam, DeepSeek, Fireworks, and OpenRouter, with credit metering and rate limiting on top of Postgres.
Engineering highlights
Single self-contained binary compiled from TypeScript via Bun, shipping natively for macOS, Linux, and Windows
OS-level sandboxing for agent commands: Seatbelt on macOS, Bubblewrap on Linux, with cycling permission modes from plan-only to full auto
A translation layer between the OpenAI Responses API and Chat Completions, so every upstream model speaks one dialect to the CLI
A per-turn hybrid router that classifies each request as fast or heavy and dispatches it to the right provider automatically
Device-code authentication pairing the CLI with the web account in seconds, no tokens pasted into terminals
Dual-currency billing (USD and INR) with credit metering at provider list cost, wired through regional payment gateways
Compliance groundwork for India’s DPDP Act: data retention policy, processor register, and a breach runbook
Product scale
Models In The Catalog
Upstream Providers Routed
OS Targets, One Binary
Want us to build yours?
Pier is what we build when we have the full stack to ourselves. If you need an AI agent system, developer tooling, or a production platform with billing built in, let's talk.