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.

What This Repo Is For

• 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.

Where To Start

Recent Heat Ranking

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: 2026-09-17 · Snapshot window: 2026-09-09 → 2026-09-17 (gain since last update, 8 days against last window’s 3, so raw gains are not comparable; every claim below is stated on the weekly rate) · 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.
• This window is 8 days against last window’s 3. Raw gains between the two columns are not comparable, so every claim here is on the weekly rate (this window: gain ÷ 8 × 7; last window: gain ÷ 3 × 7).
• Open Code Review went from off the table to #1 on a 26x jump in weekly rate, the largest this board has computed on the basis it uses for these (the previous high was 14x, K-Dense on 2026-09-01). The cause is documented, not guessed: the project hit #1 on GitHub Trending on September 16, taking +3,215 stars that day and +2,751 the day before. It grew 49% of its own size in eight days. By this board’s own rule that is a spike until a second window confirms it, and a Trending placement is the most transient cause there is.
• DeepSeek Harness accelerated 17% in its second measured window. That is the rule paying out in the other direction: two consecutive windows of growth is the confirmation this board asks for, and it now reads as a trend rather than an arrival spike. Crossed 225k.
• mattpocock/skills lost #1, down 27%. It held the seat by 11 stars last window and fell two places this one, which is what a narrow lead usually means.
• Of the two spikes this board flagged last window, one confirmed and one did not. TradingAgents added a second consecutive acceleration (+44% after +189%) and is now a confirmed trend. OpenHands gave back 32% and stayed off the table. Reporting both is the point of flagging them.
• A correction to last window’s read. This board wrote that Ruflo had confirmed rather than reversed and called it the first breakout in five windows to survive its follow-up. It then gave back 59% and left the table. The claim was made on a three-day window, which was too short a base to carry it; the pattern held for one window and not two.
• Browser Use fell 70% after entering at #5 on its first measured gain. A first measured window is not a trend either, and this board should read those entries the same way it reads spikes.
• The skills wave recovered to 3 of 10 from last window’s 2, its lowest since the wave began. addyosmani/agent-skills is back at +61%, while K-Dense (−41%) and academic-research-skills (−32%) both kept cooling.
• Milestones: Open Code Review crossed 30k and 33k, DeepSeek Harness 225k, mattpocock/skills 260k, Superpowers 285k, Hermes Agent 245k, TradingAgents and Pi both 105k, OpenCode 208k, addyosmani 95k.
• OpenClaw remains the absolute leader at 389.9k stars (+651); it is profiled but stays out of the gain-ranked table because reliable week-over-week deltas for a project this large are noisy.

Ranking Trend

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/.

Beyond The Rank

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.

Market Pulse

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 in the first week of September both ceilings moved to the same price: Claude Fable 5.1 (Sept 1) and GPT-6 Astra (Sept 3) both list at $10 / $50, so the frontier comparison is no longer about the sticker. It is about cache reads ($0.25 vs $1) and long-context shape — Anthropic bills its 1M window at flat rates, OpenAI doubles input past 272k tokens. Underneath, the tier ladders are what most work should run on: Opus 5 ($5/$25, the Claude Code default since July 24) and Sonnet 5 ($2/$10) on one side, GPT-5.6 Sol/Terra/Luna on the other, with Sol’s promotional $4/$20 running to at least Nov 21. Full table in cost & benchmarks; lineage in 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.” Since Codex CLI rust-v0.153.4 (Sept 4) the bundled default model there is GPT-6 Astra, which means the product’s default now inherits Astra’s restricted cybersecurity behaviour. See Codex.

The First Cut Of The Map

Current Mainstream Coverage

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

Example Reading Paths

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

Disclaimer

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.