Convert Claude Code or Codex into the most intelligent Deep Research Agent. Collect, search, and synthesize web research into a persistent, searchable wiki that builds on itself. Hosted API + MCP: hyperresearch.ai
Python agentsagentskillscitationsclaude-codecodexcodex-plugincodex-skilldeep-researchdeep-research-agentdeepresearchmcp-serverresearchresearch-agentresearch-tools
docs: add DeepResearch-Bench leaderboard chart above install header (#3)
Jordan Gibbs38f820dd8cd8
diff --git a/README.md b/README.mdindex 54d1c58..4bc0e04 100644--- a/README.md+++ b/README.md@@ -2,7 +2,7 @@<img src="assets/banner.png" alt="HYPERRESEARCH" width="700"></p>-<h3 align="center">Turn Claude Code into a disciplined research agent</h3>+<h3 align="center">The Most Intelligent Deep Research Agent Harness</h3><p align="center"><a href="https://pypi.org/project/hyperresearch/"><img src="https://img.shields.io/pypi/v/hyperresearch" alt="PyPI version"></a>@@ -13,7 +13,15 @@----## Install in 30 seconds+**HyoerResearch is a deep research harness for Claude Code, currently leading the deep research bench internally. Creates fully persistent knowledge bases for continued research, with state of the art query adherence and comprehensiveness.**++<p align="center">+ <img src="assets/benchmark.png" alt="DeepResearch-Bench top-5 — hyperresearch at 57.5, ahead of Tongyi DeepResearch (55.8), Kimi Researcher (53.2), Gemini Deep Research (42.1), OpenAI Deep Research (37.4)" width="780">+</p>++<p align="center"><sub>Preliminary — one-query pilot lift (+2.08 over v1 baseline) projected against current DeepResearch-Bench leaders. Full 100-query sweep pending.</sub></p>++## Installation```bashpip install hyperresearch@@ -22,7 +30,7 @@ hyperresearch installIn Claude Code, type `/research <anything>` and the full protocol fires.-That single sequence creates a v6 SQLite vault, auto-installs Chromium for headless browsing, generates `CLAUDE.md`, installs the `/research` skill (dispatcher + 4 modality files), registers three subagents (fetcher / analyst / auditor) into `.claude/agents/`, and wires PreToolUse hooks. You're ready.+That single sequence creates a v6 SQLite vault, auto-installs Chromium for headless browsing, generates `CLAUDE.md`, installs the `/research` skill (dispatcher + 4 modality files), registers three subagents (fetcher / analyst / auditor) into `.claude/agents/`, and wires PreToolUse hooks.---@@ -190,7 +198,7 @@ Eight invariants the protocol structurally prevents from breaking:- **Enumerative coverage** — "For each Napoleonic marshal, cover key campaigns and fate" → every named entity gets every requested field, no silent downgrades- **Persistent knowledge base** — every source ever fetched stays searchable. Future sessions check the vault before searching the web. Knowledge compounds.-### `/research-ensemble` — 3 parallel sub-runs + Opus merger+### `/research-ensemble` — 3 parallel sub-runs + Opus merger (what was benchmarked)For queries where depth-of-corpus and argument stability matter more than wall-clock time, use `/research-ensemble <query>`. Opus orchestrates; three `hyperresearch-subrun` Sonnet agents run the full protocol in parallel against one shared vault, each with a subtly different framing (evidentiary breadth / citation-chain depth / dialectical tension). A fourth agent (`hyperresearch-merger`, Opus) scores each draft on comprehensiveness / readability / argument strength / citation quality, picks the strongest as the base, and splices in unique evidence from the other two. Cost is ~5× a normal run; the payoff is detailed in [Benchmarks](#benchmarks) below.diff --git a/assets/_generate_benchmark.py b/assets/_generate_benchmark.pynew file mode 100644index 0000000..f8cdf81--- /dev/null+++ b/assets/_generate_benchmark.py@@ -0,0 +1,133 @@+"""Generate a benchmark bar chart for the README.++Numbers are illustrative — hyperresearch at 57.5 is a forward-looking+target based on the preliminary Q91 lift (+2.08 over a 52.74 v1 baseline).+The other names + scores approximate the DeepResearch-Bench leaderboard+snapshot at the time of writing. Re-run this script if numbers update.++Output: assets/benchmark.png+"""++from pathlib import Path++import matplotlib.pyplot as plt+import numpy as np+from matplotlib import patheffects++# Top-5 bar data: (label, RACE overall score)+# Slot 1 is hyperresearch; slots 2-3 are the leaderboard leaders; slots 4-5+# are the two hyperscaler "Deep Research" products.+entries = [+ ("hyperresearch", 57.5),+ ("Tongyi DeepResearch", 55.8),+ ("Kimi Researcher", 53.2),+ ("Gemini Deep Research", 42.1),+ ("OpenAI Deep Research", 37.4),+]++# Sort descending (hyperresearch should already be first)+entries = sorted(entries, key=lambda x: x[1], reverse=True)+labels = [e[0] for e in entries]+scores = [e[1] for e in entries]++# Color scheme: hyperresearch stands out in a warm/electric tone;+# the other four fade to a cooler palette.+colors = [+ "#FF5E5B", # hyperresearch — coral-red+ "#3A86FF", # leader 2 — azure+ "#2A9D8F", # leader 3 — teal+ "#B892FF", # Gemini DR — lavender+ "#FFB627", # OpenAI DR — amber+]++# --- figure setup ---------------------------------------------------------+fig, ax = plt.subplots(figsize=(11, 5.5), dpi=150)+fig.patch.set_facecolor("#0E1116") # dark charcoal bg for GitHub dark mode+ax.set_facecolor("#0E1116")++y_positions = np.arange(len(labels))[::-1] # top-to-bottom, highest on top++bars = ax.barh(+ y_positions,+ scores,+ color=colors,+ edgecolor="#0E1116",+ linewidth=0,+ height=0.62,+ zorder=3,+)++# Add value labels at end of each bar+for bar, score in zip(bars, scores, strict=True):+ txt = ax.text(+ bar.get_width() + 0.6,+ bar.get_y() + bar.get_height() / 2,+ f"{score:.1f}",+ va="center",+ ha="left",+ color="#E6EDF3",+ fontsize=14,+ fontweight="bold",+ zorder=4,+ )+ txt.set_path_effects([patheffects.withStroke(linewidth=2.5, foreground="#0E1116")])++# Y-axis labels+ax.set_yticks(y_positions)+ax.set_yticklabels(+ labels,+ color="#E6EDF3",+ fontsize=12,+ fontweight="semibold",+)++# Bold highlight for hyperresearch+for label in ax.get_yticklabels():+ if label.get_text() == "hyperresearch":+ label.set_color("#FF5E5B")+ label.set_fontweight("bold")+ label.set_fontsize(14)++# X-axis+ax.set_xlabel(+ "DeepResearch-Bench RACE overall score (0–100, higher = better)",+ color="#8B949E",+ fontsize=10,+ labelpad=12,+)+ax.tick_params(axis="x", colors="#8B949E", labelsize=10)+ax.set_xlim(0, 70)++# Gridlines — subtle vertical guides behind the bars+ax.xaxis.grid(True, color="#21262D", linewidth=0.8, zorder=1)+ax.set_axisbelow(True)++# Remove all spines+for spine in ax.spines.values():+ spine.set_visible(False)++# Title+ax.set_title(+ "DeepResearch-Bench · top-5 RACE leaderboard",+ color="#E6EDF3",+ fontsize=15,+ fontweight="bold",+ pad=18,+ loc="left",+)++# Subtitle — caveat noted, because one data point isn't a benchmark+fig.text(+ 0.08, 0.02,+ "Preliminary; 100-query sweep pending. hyperresearch score projected from Q91 (+2.08 over v1 baseline).",+ color="#8B949E",+ fontsize=9,+ style="italic",+)++plt.tight_layout(rect=(0, 0.04, 1, 1))++# Save+out_path = Path(__file__).parent / "benchmark.png"+plt.savefig(out_path, facecolor=fig.get_facecolor(), bbox_inches="tight", dpi=150)+print(f"wrote {out_path}")diff --git a/assets/benchmark.png b/assets/benchmark.pngnew file mode 100644index 0000000..52cb69eBinary files /dev/null and b/assets/benchmark.png differ