R28
R28 builds open-source infrastructure for AI agents that act through APIs. Its flagship Charter library lets developers declare API tools field by field, enforce authorization and validation before requests are sent, and run the calls in their own process without a proxy or per-call charge.
At a Glance
- Developers and teams building AI agents
- SaaS companies whose agents act in end-user accounts
- Security and platform teams reviewing agent access and API egress
- Open-source and self-hosted agent/MCP users
- +1 more
AI Tools by R28
(1)
Charter
Python Library for Agent API Validation
Discussions
No discussions yet
Be the first to start a discussion about R28
Latest News
Charter 0.6.0 released, the latest release listed by the project.
Charter 0.5.2 released with further fixes and updates documented in the changelog.
Charter recipes added demonstrations for thread size, field cuts, glosses and schema cost across Gmail, Google Docs, Linear and Stripe.
Charter 0.5.0 and 0.5.1 releases expanded OAuth, credential and authorization handling.
Products & Services
An open-source Python library for agents acting in customer accounts through APIs. It represents API tools as Pydantic models, validates every call before sending, supports per-call account context and egress maps, generates deterministic wire formats, and includes ready-made packs for multiple APIs.
Small runnable Apache-2.0 demo scripts built on Charter, covering Gmail, Google Docs, Linear and Stripe tasks such as reply-all, attachment saving, schema-cost analysis, field cuts and validation glosses.
An open-source local example app for OpenAI Dots and Grok Bot. It uses the user's own model and connected-app keys, reads email, calendar, GitHub, Slack and Linear, and asks before making changes; the repository explicitly says it is maintained as an example rather than a product with a roadmap.
An MCP server with ten workflows for competitor monitoring, cited research, paper alerts, market maps, due diligence and related work, integrating Firecrawl, Tavily, Notion, Slack, Google Docs/Sheets, GitHub and Linear.
Market Position
R28 positions Charter as typed, declarative control and wire-format infrastructure rather than another agent framework or hosted API proxy. It contrasts with hand-written tool glue and API-specific or popular MCP servers: Charter emphasizes field-level authorization, deterministic request construction, preflight refusal and reviewable egress maps, while remaining compatible with agent frameworks such as OpenAI and LangChain and MCP clients.
Leadership
Founders
Nathan Queme
Founder of R28 and builder of Charter. He previously worked as a backend developer at QSSTUDIO and as a freelance full-stack developer; he also built SurfBuddy solo, running agents in real Gmail, Calendar, Docs and Sheets accounts.
Executive Team
Nathan Queme
Founder
Founder and primary builder of R28/Charter; previously a backend developer at QSSTUDIO and a freelance full-stack developer, and creator of the solo-built SurfBuddy agent project.
Founding Story
Nathan Queme says he began building agents in real Gmail, Calendar, Docs and Sheets accounts in 2024 through SurfBuddy. Charter grew out of the hardened plumbing needed to control what those agents could see, send and receive in each API, and was then released as open-source Apache 2.0 software.
Business Model
Revenue Model
Charter is distributed as open-source Apache 2.0 software and runs in the user's own process. The official materials state there is no proxy, no per-call pricing and no telemetry; the public materials do not describe a paid subscription or hosted-service tier.
Target Markets
- Developers and teams building AI agents
- SaaS companies whose agents act in end-user accounts
- Security and platform teams reviewing agent access and API egress
- Open-source and self-hosted agent/MCP users
- Research and operations teams using connected-app workflows
- SaaS products adding agents that operate inside customers' Gmail, Stripe, Notion or other accounts
- Secure customer-account automation with explicit tenant/account boundaries
- Building and reviewing tool surfaces for coding agents and AI assistants
- Replacing hand-written HTTP or MCP tool glue with typed declarations and preflight validation
- Research, competitor monitoring, due diligence and market mapping through the Web Research to Docs MCP server
- Local personal assistants that connect email, calendars, GitHub, Slack and Linear