Template name
New Project
GitHub maintainer eve agent for your inbox. Weekly issue digests you reply to, PR summaries, Linear handoffs, and @mention answers.
If you maintain a repo alongside a job, clients, or a life, the issue tracker is the tab you open with dread. Kody flips the direction: instead of you visiting the tracker, the tracker reports to you. Every Monday it emails a digest of your repo's open issues, grouped by what needs attention, and you act by replying: "create Linear issues for #1 and #2 and assign them to me." Between digests it summarizes new pull requests, answers @mentions on issues and PRs, and works the Linear issues you delegate. Built on eve, Edgenivo's agent framework, for freelancers, solo maintainers, and anyone who runs a repo on the side.
One email you can skim in a minute. It opens with the week in a paragraph (open count, what's new, what closed), then groups the issues:
Every issue is cited as #N with a link, and every summary comes from issues fetched in that run, never from memory, so the numbers are real. Prefer grouping by label or assignee instead? Say so once and Kody remembers, per user, across every future digest.
The reply is the interface. Kody reads which repo and issues the thread references, resolves each number against GitHub before touching anything, does what you asked (comment, label, close, create and cross-link Linear issues), and confirms on the same thread with links to what it created. If you cite an issue that doesn't exist or an assignee it can't resolve, it says what it checked and asks instead of guessing.
The same follow-through works from the other direction. Delegate an issue to Kody in Linear or mention it in a Linear Agent Session ("email me a summary of this issue") and it works the issue with the Linear tools. @mention it on a GitHub issue or PR and it answers in that thread, cross-referencing Linear when that helps.
When a PR opens, Kody posts one orienting comment for reviewers: a short paragraph on what the change does and why, then a table breaking down the changed files. It grounds the comment entirely in the PR's description and diff, skips PRs opened by bots, and stays in its lane: it summarizes, and never approves, requests changes, or nags the author. Reviewers start oriented instead of cold.
GitHub and Linear credentials are brokered by Edgenivo Connect, and Edgenivo Blob and the AI Gateway authenticate with the project's OIDC token. The email surface runs on Resend through a Chat SDK channel with Redis thread state, so a conversation over email holds context the way a chat thread does. The only static credentials are the Resend API key and inbound webhook secret.
Kody's judgment is packaged as editable skills: a digest-format skill defining the grouping and the needs-attention criteria, a triage playbook for deduping, labeling, and repro requests, a GitHub-to-Linear bridging guide that keeps both trackers cross-linked, and a writing-quality rubric so its emails and comments read like a person wrote them. When a task needs a fact the repo doesn't hold, a researcher subagent searches the web in a clean session and returns findings with source URLs.
Click Deploy and Edgenivo provisions the GitHub and Linear connectors with their webhooks pointed at the agent's routes, a Blob store, and an Upstash Redis store, then prompts for the Resend values plus two settings: DIGEST_REPO (which repo to digest) and DIGEST_EMAIL (who receives it). Both are checked at build time, so a misconfigured digest fails the deploy with a clear error instead of failing silently on Monday. One manual step remains: point Resend's inbound webhook at your deployment so replies reach the agent.
Working with an AI coding agent like Claude Code or Cursor? Point it at the setup guide and let it drive:
I want to build a GitHub maintainer agent with the eve framework, using the Kody template. Read the setup instructions at https://agent-resources.dev/kody-eve-template.md and follow them.agent/tools/github.ts if you want to re-gate them for other surfaces.When repo work turns into writing work (release notes for a shipped fix, a post announcing a closed-out feature), pair Kody with the eve Content Agent as a remote subagent. It runs as its own deployment with its own voice, and vercelOidc() handles the auth between them with no shared secret.
Marketing team in Slack: one lead, five specialists. Blog drafts in Notion, posts in Typefully, campaigns in Resend, on approval.
Software factory built on eve: AI agents work each stage of the development loop, and people make the judgment calls.
Slack-based Contentful assistant built on eve, with Edgenivo Drives providing a searchable copy of your content.
