Skip to content

Rebuild agent-memory.dev as a technical showcase #790

Closed

Description

@rohitg00

Why

The current agent-memory.dev reads like sales copy. It sells the product but doesn't show what agentmemory actually does on the inside:

  • iii primitive composition (workers + functions + triggers)
  • The retrieval stack (hybrid BM25 + vector + graph + reranker)
  • Multi-tier memory pipeline (observations → compressed → memories → semantic → procedural → crystals → insights → lessons → facets)
  • The iii-console live state view
  • Hooks across 12 events, 4 skills, 18 MCP tools, 49 REST endpoints
  • 17 connect adapters
  • Real-time observation stream, dashboard, graph visualisations from the viewer

We want a site that earns a developer's trust by showing the engineering instead of pitching the outcome.

Brief

Build a new site for agentmemory. Anything goes on framework + stack as long as it deploys statically or as a simple SSR app. The site should feel closer to a deep-tech product page than a SaaS marketing page.

Reference vibe (none are direct copies — they're examples of "technical pages that feel alive"):

  • stripe — tight, technical, animated transitions on every section
  • tldraw.dev — embeds the live product
  • astro.build — feels like the docs are part of the landing
  • linear.app — pacing + scroll-tied animation

Process

  1. Fork rohitg00/agentmemory.
  2. Build the website on your fork — feel free to use a separate website-redesign/ directory, separate branch, or a sibling repo if it's cleaner.
  3. Deploy it somewhere public (Vercel / Netlify / Cloudflare Pages / GH Pages all fine).
  4. Open a PR back to this repo. The PR body must include:
    • Live demo URL (deployed somewhere reachable)
    • GIFs of the key interactions (autoplay short loops, ~5-10s each)
    • Screenshots of the major sections (light + dark mode if applicable)
    • Short walkthrough video (~60-90s; Loom / YouTube unlisted / mp4 in PR is fine)
    • One paragraph on the design rationale
    • Stack used + a brief why
  5. The strongest submission becomes the official agent-memory.dev — credit to the author preserved in the repo + README.

What we want to see (judging axes)

Roughly weighted:

Axis Weight What it means
Technical depth 35% Shows the real architecture — workers, triggers, retrieval, tiers — not a feature list
Interactive proof 25% Live or near-live demos: query the embedded engine, render a real graph, animate a real consolidation flow. Static screenshots count for less than a working playground.
Performance 15% Lighthouse 95+, no jank, no 5MB hero bundle, fast on mobile
Aesthetic 15% Editorial typography, restrained motion, real grid. No SaaS stock illustrations.
Accessibility 10% Keyboard + screen-reader paths actually work; motion respects prefers-reduced-motion

Sections we'd like to see covered

  • Hero that shows agentmemory in motion (not a 3D blob)
  • "What is iii" composition diagram — workers / functions / triggers as primitives
  • Memory tiers, with a real example flowing through each stage
  • Live (or near-live) hybrid search demo against a seeded corpus
  • Knowledge graph visualisation (use the viewer's real layout if possible)
  • iii-console snapshot — live event stream, state mutations, audit log
  • Connect adapters grid (17 today: Claude Code, Cursor, Codex, Copilot CLI, Cline, OpenCode, OpenClaw, Hermes, pi, Gemini CLI, Kiro, Antigravity, Qwen, Continue, Crush, Factory, plus the standalone MCP shim)
  • Pricing-free comparison vs Letta / mem0 / zep — feature axes, not bullet wars
  • Quickstart that drops a copy-paste command and shows the agent doing recall in under 10 seconds
  • Docs landing that doesn't feel grafted on

Constraints

  • No emojis in headings or hero copy (the rest of the repo follows this).
  • No em-dashes (we strip them on sight; ASCII hyphen + space is fine).
  • Brand palette is up to you, but the wordmark stays as the existing SVG.
  • License: contributions accepted under the repo's existing license. Use no proprietary illustrations / fonts you don't have rights to.
  • No competitor logos in mockups without permission.
  • Real numbers in the stats grid only. If you cite "53 MCP tools / 125 REST endpoints / 1300+ tests" the number has to be auto-derived (script that greps the source). The repo does this today for the existing site; see website/scripts/gen-meta.mjs.

What stays / what changes

  • The current site at website/ can stay as a reference. We're not deleting it during the contest.
  • The new winner will land in a fresh top-level website-v2/ (or whatever the author names it), then we swap the deploy target.
  • Existing routing (agent-memory.dev, agentmemory.dev) will point at the new build once selected.

Recognition

  • Author keeps a Built by @<handle> attribution in the repo and in the site footer.
  • Listed as the official designer in the README + release notes.
  • Permanent commit history with their name on the merged PR.

Open questions / answer in your PR

  • Framework choice + why (Next? Astro? Vite + React? Static HTML?)
  • How you'd ship updates to the tier-by-tier walkthrough as we add new tiers
  • How you'd handle the live demo's seeded data — bundled JSON snapshot, server-side, WebContainers?

Submissions

Comment with a link to your fork + WIP URL as soon as you start, so we can avoid two people redoing the same section. No deadline — rolling submissions. Selection happens roughly when the third or fourth credible submission lands, so we have something to compare against.

If anything in the brief is ambiguous, ask here before you build — better than a polished site we can't merge for license / brand reasons.

Activity

  1. pinned this issue on Jun 2, 2026
  2. changed the title [-]Community design contest: rebuild agent-memory.dev as a technical showcase[/-] [+]Rebuild agent-memory.dev as a technical showcase[/+] on Jun 2, 2026
  3. Repository owner deleted a comment from stevesagronegocios673-ux on Jun 2, 2026
  4. MannXo commented on Jun 3, 2026

    @MannXo

    Can I take this? I have some time to work on it this week.

  5. crazy0codes commented on Jun 3, 2026

    @crazy0codes

    I would also like to work on this

  6. sham-jadhav03 commented on Jun 9, 2026

    @sham-jadhav03

    I have some clarification questions.

    1. Must demos use the real AgentMemory runtime, or are deterministic simulations acceptable?
    2. Can contributors reuse internal graph/viewer components, and where are those APIs documented?
    3. Will judging use a published scorecard, or is selection ultimately subjective?
  7. rohitg00 commented on Oct 8, 2026

    @rohitg00
    OwnerAuthor

    Done in v0.9.30: agent-memory.dev rebuilt around use cases and sourced proof, with benchmarks, security, privacy, changelog and comparison pages, robots.txt, sitemap, llms.txt and llms-full.txt. https://github.com/rohitg00/agentmemory/releases/tag/v0.9.30

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions