updated 2026-08-06 05:00 UTC

AgentBench

THUDM/AgentBench

A Comprehensive Benchmark to Evaluate LLMs as Agents (ICLR'24)

Observability & Evaluation Python Grade D Listed Apache-2.0
34.8/100
Rank #399 of 468
Compare AgentBench

How does it stack up against its Observability & Evaluation neighbours?

Pick any agent to compare →

Thin or incomplete trust evidence. Review carefully before production use.

Open compare tool Suggest correction
Listing state
Listed
Evidence coverage
Grade C · 2/5 signals
Last push
2026-02-08 · 179d ago
Recent change
No recent signal change

Is AgentBench safe? AgentBench scores 34.8/100 (Grade D), ranked #399 of 468 tracked open-source AI agent projects, on evidence coverage C (2 of 5 independent signal types). The public evidence: no package-provenance attestation found; OSSF Scorecard rates its supply-chain practices 3.4/10; 51% of recent commits are signed; last pushed 2026-02-08. Every point is earned from checkable signals — never paid placement. How scoring works →

Ranked neighbours in Observability & Evaluation

#17 Opik — 46.5 (D) → #18 Ragas — 45.2 (D) → #19 TruLens — 41.4 (D) → #20 Webarena — 35.4 (D) →

Quick Trust Read

What Would Improve It
Publish package provenance or release attestations for stronger supply-chain evidence.
Recent Changes
2026-07-29
Rank Moved
Rank dropped 11 spots (#389 → #400)
2026-07-15
Rank Moved
Rank dropped 27 spots (#354 → #381)
2026-07-14
Rank Moved
Rank dropped 20 spots (#334 → #354)
Maintainer Checklist
Raise Scorecard signals Current OSSF Scorecard is 3.4/10. Tighten the weakest checks to improve public safety evidence.
Publish provenance Add package provenance or release attestations so users can verify where shipped artifacts came from.
Refresh maintenance signals The repo was last pushed 179 days ago. Fresh activity helps separate stable projects from stale ones.
Compare with another agent Get email alerts for AgentBench View Observability & Evaluation Report missing data
33.0
Activity sub-score · out of 100
#21
In Observability & Evaluation

How to read this: HVTrust (0–100) weighs supply-chain signals (provenance, OSSF Scorecard, signed commits, open license) alongside real-world adoption. Grade D reflects the trust score band: A ≥ 80, B ≥ 65, C ≥ 50, D < 50. Evidence coverage C is separate — it grades how many independent signal types back the score (2 of 5), so a high score on thin evidence stays visible. Full methodology →

Signals refreshed 2026-08-06 05:00 UTC · Repo last pushed 179 days ago

Rank Trend

2026-07-05 2026-08-06

Activity & Reach

Stars
3.6k
Forks
272
Last Push
2026-02-08
179 days ago
Commits (4 wk)
0
Downloads (7d)
—
HN mentions (30d)
—
Open Issues
63
Rank Change
=
was #399

Analysis

HVTrust Dimensions vs Observability & Evaluation

34.8 / 100 · 100.0% confidence

AgentBench Observability & Evaluation average (21 agents)

Safety / Integrity50% OSSF Scorecard · 30% provenance · 20% signed commits
6.8 / 25
4.5 below avg 11.3
Identity / Provenance60% listing status · 40% build provenance
10.8 / 18
2.4 below avg 13.2
Transparency50% declared license · 50% OSSF Scorecard
11.4 / 17
0.5 below avg 11.9
Maintenance60% last-push freshness · 40% commit activity
0.1 / 20
15.0 below avg 15.1
AdoptionLog-scaled stars · package downloads
8.5 / 20
6.1 below avg 14.6

Activity Inputs

33.0 / 100
StarsRepository reach
21.4 / 30
FreshnessLast push recency
0.1 / 25
ActivityRecent commits
0.0 / 25
CommunityFork signal
11.3 / 20

Supply Chain Trust

Package Provenance
None
No package attestations found
OSSF Scorecard
3.4 / 10
OpenSSF Scorecard · scanned Aug 5, 2026
Signed Commits
51%
of last 100 commits verified
Binary-Artifacts 10
Branch-Protection 6
CI-Tests 0
CII-Best-Practices 0
Code-Review 6
Contributors 10
Dangerous-Workflow 10
Dependency-Update-Tool 0
Fuzzing 0
License 10
Maintained 0
Packaging -1
Pinned-Dependencies 0
SAST 0
Security-Policy 0
Signed-Releases -1
Token-Permissions 0
Vulnerabilities 0

Is AgentBench safe?

Public trust evidence for AgentBench is thin: several supply-chain signals are missing or weak. This does not mean the project is unsafe — it means an outside observer cannot easily verify the usual integrity checks. Treat with extra scrutiny.
Does AgentBench publish package provenance?
No published build provenance is currently detected for AgentBench. This is common for open-source projects but means consumers cannot independently verify that the package on the registry matches the GitHub source.
Does AgentBench have an OpenSSF Scorecard?
AgentBench has an OpenSSF Scorecard score of 3.4/10. The Scorecard checks for branch protection, signed releases, dependency updates, fuzzing, code review, and other supply-chain hygiene items. See the full check breakdown on this page.
Is AgentBench actively maintained?
Slowing down. Last push was 179 days ago — keep an eye on whether activity resumes.
What license does AgentBench use?
AgentBench ships under Apache-2.0. A declared, OSI-approved license is one of the transparency signals HVTrust scores.
Are AgentBench's commits signed?
50% of the last 100 commits to AgentBench are verified-signed (GPG, SSH, S/MIME, or GitHub's signing flow). Signed commits help confirm that code was authored by who the commit claims.

Not a safety endorsement. HVTracker describes what public signals show, not whether a project is safe for your use case. Run your own security review before adopting in production.

AI agent surface

MCP, providers, tool surface
Scored in HVTrust

These runtime-trust fields — detected from public repo docs and manifests — contribute a bounded adjustment to this project's HVTrust score alongside supply-chain evidence. The exact values each field can add or subtract are documented in the methodology → Compare this surface across every listed agent in the capability matrix →

MCP Server Support
None detected
No MCP server signal detected.
Detailed evidence is not shown in the public view.
External Service Dependencies
high confidence
3 detected
Public provider/service dependencies detected.
Credential signal: No explicit API-key/config marker detected.
Tool / Plugin Surface
high confidence
Extensions
Extension based plugin/integration surface detected.
  • browser
  • code
  • database
  • shell
Detailed evidence is not shown in the public view.
Package Provenance Drift
N/A
No package source configured
Detailed evidence is not shown in the public view.
  • MCP signal live
  • External deps live
  • Tool / plugin surface live
  • Package provenance drift live

Maintain AgentBench?

For maintainers

HVTrust scores AgentBench from public signals only — we never contact maintainers first. If a signal is wrong, stale, or missing (provenance you publish, a Scorecard you run, signed releases), tell us and we'll review it. Corrections are public and tracked on GitHub.

Submit a correction → Add the trust badge

Reputation Timeline

Signal history
Rank 6Delisted 1Listed 1HVTrust 1Scorecard 1
2026-07-29
Rank Moved
Rank dropped 11 spots (#389 → #400)
2026-07-15
Rank Moved
Rank dropped 27 spots (#354 → #381)
2026-07-14
Rank Moved
Rank dropped 20 spots (#334 → #354)
2026-07-13
Rank Moved
Rank dropped 36 spots (#298 → #334)
2026-06-24
Rank Moved
Rank dropped 35 spots (#250 → #285)
2026-06-23
Scorecard Added
OSSF Scorecard: 3.4/10
2026-06-23
Rank Moved
Rank rose 48 spots (#298 → #250)
2026-06-23
HVTrust Changed
HVTrust up 23.8pts (16.7 → 40.5)
2026-06-21
Newly Listed
First tracked at rank #297
2026-05-25
Removed From Active Tracking
Removed from active tracking
GitHub →

Embed Badge Badge guide for maintainers →

For maintainers
HVTrust 34.8 Grade D
Markdown:
[![HVTrust](https://hvtracker.net/badge/agentbench.svg)](https://hvtracker.net/agents/agentbench)
HTML:
<a href="https://hvtracker.net/agents/agentbench"><img src="https://hvtracker.net/badge/agentbench.svg" alt="HVTrust"></a>

Other agents in Observability & Evaluation

Data sources
GitHub REST API (repo, commits, stars, forks, license) · OpenSSF Scorecard CLI
Each agent's signals refresh once daily across 6 staggered batches. Methodology v4.2 · Raw JSON