Greg Devo
Greg Devo is an independent developer and the creator/maintainer of Memdoor, an open-source CLI coding agent written in Go. Memdoor aims to let developers run coding agents and recoverable multi-step workflows locally, over SSH, in CI, on schedules, in editors, or from a phone, using the model provider or local model of their choice.
At a Glance
- Individual software developers
- Open-source maintainers
- Engineering teams using CI and scheduled automation
- Enterprise engineering organizations requiring self-hosting and approval controls
- +2 more
AI Tools by Greg Devo
(1)Memdoor
Open Source CLI Coding Agent
Discussions
No discussions yet
Be the first to start a discussion about Greg Devo
Latest News
Memdoor v1.814.0 released with one binary per platform and SHA256-verified installation assets
Greg Devo reported 92 of 100 LiveCodeBench problems solved with DeepSeek V4.1 Flash using the Team Mode approach
Memdoor Team Mode was featured on Hacker News: five Qwen3.8-27B agents averaged 9.8/10 on the first ten LiveCodeBench Hard problems
Greg Devo published lessons and results from running SWE-bench with a team of DeepSeek V4.1 Flash agents
Products & Services
Open-source CLI coding agent in one Go binary. It reads and edits project files, runs commands and tests, supports Team Mode, and exposes the same gateway through a terminal window, headless CLI/CI, cron, editors, SSH, and a phone remote.
Declarative or agent-authored multi-step graphs for coding work. Independent steps can run in parallel; each step has a target/proof, failed runs can resume, and approval gates let a person review before continuation.
Memdoor's multi-agent mode, activated from the CLI, in which several fresh-context agents such as planners, ideators, solvers, and managers communicate through files and a checking/review loop selects or verifies the result.
Greg Devo's related open-source local-LLM inference project in Go, described on his GitHub profile as an OpenAI-compatible server for MLX on Apple silicon, ggml on Linux/CUDA, or pure Go.
Market Position
Memdoor positions itself as an open-source, self-hostable CLI coding agent and workflow engine that combines a single gateway, verifiable multi-step workflows, Team Mode, local-model support, and broad access doors. Its own comparison page names Claude Code, Codex CLI, Gemini CLI, Aider, OpenCode, Cline, Copilot CLI, Amp, Goose, Cursor CLI, Command Code, Kilo Code, pi, and omp as comparison points; its differentiation is particularly the workflow graph/proof model, local or BYOK operation, and phone/SSH/editor access through one gateway.
Leadership
Founders
Greg Devo
Independent developer; GitHub profile describes him as 'Ex Spotify' and says he is building the next generation of open-source AI harnesses. He maintains Memdoor and related open-source Go projects including Janis and Mario.
Executive Team
Greg Devo
Creator and maintainer
Independent developer and former Spotify employee; maintains Memdoor and its public GitHub repositories.
Founding Story
In a personal post, Greg Devo wrote that he started Memdoor in June 2026 with a small ambition: a lean coding agent in the terminal, written in Go, shipped as one binary, and able to run on whatever model key he already paid for. The project developed around making inexpensive or local models more capable through workflows, multiple fresh contexts, file-based communication, and verifiable checks.
Business Model
Revenue Model
The core product is free and open source under Apache-2.0. Users bring their own model provider key or connect a supported subscription; model usage is paid to the relevant provider at list price. Memdoor also offers a supported, self-hosted Enterprise service by invoice.
Pricing Tiers
Apache-2.0 software; self-host and run on a laptop or company machines. No account is required. Users pay their chosen model provider separately when using a provider key.
Supported self-hosting on company machines, gateway setup, named-person support, approval mode, tool guards, and company AI-gateway/vendor-key operation.
Target Markets
- Individual software developers
- Open-source maintainers
- Engineering teams using CI and scheduled automation
- Enterprise engineering organizations requiring self-hosting and approval controls
- Developers and researchers working with local or open-weight models
- Data/automation teams that need durable workflow state and retries
- Interactive software development in a terminal
- Automated issue-to-pull-request and code-review workflows
- CI jobs and headless scripting where the exit code is the receipt of checks
- Scheduled repository maintenance and unattended backlog processing
- Running coding agents on a remote machine or over SSH
- Local/private coding assistance using open-weight models