AI Agent Map

AI Agent Map is a practical, visual-first guide for comparing mainstream AI agents, agent platforms, runtimes, and orchestration tools.

The goal is simple: help readers get to a sensible shortlist faster.

• The agent landscape is crowded.
• Many resources explain ideas, but not fit, anti-fit, or operating cost.
• People usually need a comparison layer, not another pile of links.

This repo stays focused on selection: what a system is good at, where it breaks down, and what kind of operator cost comes with it.

If your question is...	Start here
I need a shortlist first
I need help choosing for coding automation
I already have candidates and want a side-by-side view
I care about dimensions like approval, memory, scheduling, and deployment
I want every project scored on those dimensions, side by side	Capability matrix
I want to know what it actually costs to run, and which model tier is worth it	Cost & benchmarks · Memory approaches
My agents already run — I need to know whether they still work	Observability & evaluation
I want the stock rankings and the weekly trend chart
I want problem-first guides or the full comparison list	Use cases · Comparisons

Popularity is not fit.

This table tracks projects that showed up as especially hot in the latest weekly GitHub snapshot. The rank follows the 7-day gain. The total star counts below were checked when this repo was updated.

Last updated: 2026-08-27 · Snapshot window: 2026-08-12 → 2026-08-27 (gain since last update, 15 days — two scheduled Wednesday refreshes were missed, so this is a catch-up window; approximate) · Star counts: checked at update time

Project names link to the upstream GitHub repo. When this map has a written profile, it is linked separately in the "Map status" column.

• Heat is useful for discovery, not for selection by itself.
• Two refreshes were missed, so read every number here as 15 days, not 7. The Wednesday updates for 2026-08-19 and 2026-08-26 did not run; this window spans 2026-08-12 → 2026-08-27. Raw gains are therefore roughly 2.1× a normal window and are not comparable to last week's column. Every "up"/"cooled" claim below is stated on the weekly rate (gain ÷ 15 × 7), which is comparable; the table's own numbers are the raw 15-day gains.
• Codex CLI is the story, and it is a vendor story. It went from #9 to #2 on +13,321 — a weekly rate of ~6.2k against ~1.4k last window, a 4.5× re-acceleration and the largest this board has recorded. It sits on two dated events inside the window: OpenAI cut GPT-5.6 Sol from $5/$30 to $4/$20 per M tokens on Aug 21 for three months, applying to Codex credits as well as the API, and reported 20M active Codex users the same day. This is the first time the heat board's top mover is explained by a price change rather than a launch.
• mattpocock/skills takes a tenth straight #1 and re-accelerates (237.9k, +23,585, ~11.0k/week against 10,291). The single deceleration flagged last window did not continue. It still out-gains everything below #4 on this table combined.
• Last window's read on addyosmani/agent-skills was a spike, not a re-acceleration. The "sharpest single-window re-acceleration this board has recorded" gave back 64% of its weekly rate (4,664 → ~1,698/week) and fell from #2 to #6. Two windows running, this board has now over-read a one-window jump — jcode in August, addyosmani here. The pattern is worth stating as a rule: a single 4–5× window is a spike until a second window confirms it.
• TradingAgents crossed 100k and is still not in scope (100,735). It has now been on this table four windows running purely on gain. The line at the top of this section does the work: this table ranks heat, not fit.
• The middle of the board is one broad cool-down. Six of the top ten posted a lower weekly rate than last window, and outside the top ten the pattern is sharper: QM gave back 71% of its rate (1,756 → ~501/week, off the table from #8), Open Code Review 53%, jcode another 50%, AutoGPT 77% — last window's "9× wake-up" was noise. Against that, Pi, n8n, Claude Code and Grok Build all improved.
• n8n is back one window after losing its seat, and OpenHuman (+1,955, 38.2k) missed the table by 259 stars — its best showing since being profiled.
• New-inclusion decision: one profile added — eve (vercel/eve, Apache-2.0, 4.8k). A backfill rather than a breakout: Vercel shipped it at Ship London on June 17 2026 as part of its Agent Stack, and this map missed it. It belongs on the build-your-own route and it changes that route's shape — an agent is a directory of files, and durable execution, per-agent sandboxes, needsApproval gates, subagents, evals, and eight-plus channel adapters ship inside the framework. It is the first entry on that route with real delivery surfaces, and the only one whose documented production path runs through one vendor's platform. Also scanned and held on the watchlist: truefoundry/trueforge (4.6k, MIT, "the runtime layer that turns an LLM into a working agent"), trailhq/Graft (4.9k, MIT, a shared code-graph context layer under Claude Code / Cursor / Codex / Gemini) and fuxicodex/Fuxi (2.1k in three weeks, but no clear license — NOASSERTION, which is a hard blocker for a profile here). Carried over and still growing: Vincentwei1021/video-shotcraft (6.4k), QwenLM/Qwen-MM-Plugins (2.8k).

How the weekly top 10 has shifted since tracking began — each line is one project, breaks mean it fell off the board that week:

And the same windows read as seats per layer — the quantitative version of the skills-wave story the bullets tell in prose:

Full stock rankings by category — agents, agent infra, skills, and their verticals, sorted by total stars — live in rankings/.

Popularity tells you what to look at. These four pages tell you what to pick:

• Capability matrix — every project scored side by side (●/◐/○/—) across the nine shared capability dimensions, grouped by route. The answer to "for this capability, who treats it as a core strength."
• Cost & benchmarks — frontier-model capability vs per-token price, plus how each coding agent actually bills. Since the model layer went tiered and metered, "which tier for this task" is the selection decision.
• Memory approaches — six different things projects mean by "has memory," from self-editing stores to passive semantic recall, and which to pick for what you need to persist.
• Observability & evaluation — the layer under everything above: once an agent runs unattended, failure stops looking like a crash and starts looking like silent quality drift. Compares Langfuse, Opik, Phoenix, Helicone, LangSmith and others — and untangles the four different things "open source" means in that field.

The three structural stories shaping selection right now — full records with dates and sources live in market-events.md:

• The .claude/skills wave keeps compounding — and is now concentrating into one directory (May 2026 → ongoing): curated skill collections and skills frameworks have held roughly half of the weekly heat top 10 for three months, and through August the count stopped moving entirely — four of ten, three windows running, with no rotation in the last one. What is still moving is the split inside the wave: mattpocock/skills now out-gains the other three combined, where a month ago it was level with them. For many tasks the skill layer now matters as much as the underlying agent; read the concentration as key-person risk, not a broadening ecosystem. This map profiles the framework end through Superpowers and tracks collections on the skill boards.
• The model layer became a budget decision, and the budget now moves: Anthropic's Mythos-class Claude Fable 5 (June 9) sits above Opus 4.8 on metered credits, while OpenAI's GPT-5.6 (July 9) ships in three price tiers. As of August 21 the top tier is also promotional — Sol cut to $4 / $20 for three months, covering Codex credits — which put the index leader below Opus 4.8 on output price and made cost & benchmarks a document with a review date on it. Spring reference point: GPT-5.5.
• Product boundaries are collapsing upward: OpenAI merged Codex into the ChatGPT app (July 9) — on the OpenAI side, "which coding agent" is turning into "how you use ChatGPT." See Codex.

61 profiled projects, grouped by what they are. Expand a group, or browse the full route/coverage tables in agents/.

If you are still deciding where to begin, use one of these quick routes and then branch out.

Star counts and 7-day gains are point-in-time GitHub snapshots taken when the repo is updated; numbers shift quickly between weekly refreshes and small rounding differences are expected. Project descriptions, vendors, and capability summaries reflect public information at the time of writing and may change as projects evolve, get acquired, or pivot. This map is selection guidance — not endorsement, financial advice, or a production-readiness guarantee. Verify against each project's own docs before committing to a choice.