Agent Development Kit (ADK) by Google

# Fix list: Agent Development Kit (ADK)

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/google-adk, the October 2026 research run, assessed 1 October 2026. Grade BB, 74.9 out of 100.

This is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public.

For a coding agent working on Agent Development Kit (ADK): work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published.

## 1. Payments & pricing, 60 out of 100, up to 5 more on the total

Why it scored 60: No payment protocol (0). Scored on Agent Runtime, the hosted option, which publishes per-unit prices without a login ($0.085 a vCPU-hour, $0.009 a GiB-hour, 50 vCPU-hours and 100 GiB-hours free a month) (20). The Apache-2.0 package installs with no card (20) and no account, and runs local models (20).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):

The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).

- 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which.
- 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for "contact sales" or prices behind a login.
- 20, a free tier or trial that doesn't need a card.
- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).

Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.

Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol.

## 2. Agent ergonomics, 70 out of 100, up to 4.9 more on the total

Why it scored 70: An agent with one MCP server is about 15 lines with the built-in McpToolset, and tool_filter limits which tools load, with no dynamic filtering or deferred loading that we saw (20). RunConfig caps model calls per run; we didn't find context compaction in the pages we loaded (10). No exception reference or MCP error handling documented (5). Invocations are resumable and model calls take retry options (20). An agent needs a name, a model and an instruction, in Python, TypeScript, Go, Java or Kotlin (15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):

- 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).
- 20, pagination, filtering and output-size controls.
- 20, actionable, documented error responses, codes and messages an agent can recover from.
- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.
- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.

Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs.

## 3. Reliability, 80 out of 100, up to 4 more on the total

Why it scored 80: Official packages on PyPI (Requires-Python >=3.10, constraint files for 3.10 to 3.14) and npm (20). The continuous-integration workflow passes on main (25). 300 open issues and 261 open pull requests (12). The changelog calls out breaking changes, but they shipped in minor releases of a post-1.0 package (2.6.0 on 2026-07-29 and 2.7.0 on 2026-08-13), and 2.8.0 reverted an A2A guard that had broken every tool confirmation (8). 2.10.0, past 1.0 (15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):

Hosted APIs, MCP servers, models and platforms.

- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).
- 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.
- 15, rate limits documented with numbers.
- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.
- 10, an SLA published for any paid tier.
- 10, the surface agents use is generally available, not beta or preview.

Local packages, SDKs, frameworks and stdio MCP servers.

- 20, installs from an official package with supported runtimes stated.
- 25, a public CI and test suite, passing on the default branch.
- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).
- 15, semver discipline and breaking changes called out in a changelog.
- 15, version 1.0 or later, or declared stable.

Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors.

## 4. Transparency & trust, 70 out of 100, up to 2.6 more on the total

Made of editorial 67, provenance 72.

Why it scored 70: Apache-2.0 (30). Google's general privacy policy applies and the observability docs say content capture is opt-in, but there's no ADK-specific statement of what, if anything, leaves the machine (15). Breaking changes are dated in the changelog and 1.x still gets releases, with no written support window (12). We found no telemetry statement either way on adk.dev (10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):

- 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.
- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).
- 0 to 20, a deprecation policy or notices with dates.
- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).

The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two.

Provenance checks not met in full (half of this category, computed from checked facts):

- Domain age: adk.dev, registered 2019-02-28 (7 years) (11 of 15)
- Status page: not found (0 of 10)
- security.txt: not found (0 of 10)

## 5. Schema & documentation, 87 out of 100, up to 2.1 more on the total

Why it scored 87: An API reference on adk.dev (25). llms.txt with about 250 entries and Markdown versions of each page (10). The docs explain the agent types and when to use workflow agents, but say little about when not to use ADK (15). Tools are typed functions, and McpToolset keeps the server's schemas (12). Examples throughout, but the MCP page has no error handling section and we found no exception reference (10). Dated changelog and release notes (15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):

APIs and MCP servers.

- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).
- 10, llms.txt or Markdown docs served for agents.
- 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.
- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.
- 0 to 15, examples and documented error responses.
- 15, versioning and a public changelog.

Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference.

## 6. Security & auth, 92 out of 100, up to 1.4 more on the total

Why it scored 92: We found no usage telemetry in the docs, and capturing message content in traces is opt-in (30). Tool confirmation for human approval, the docs say to always pass tool_filter to McpToolset, and sandboxed code execution is recommended for model-written code (20). Before-tool callbacks, plugins, a Model Armor plugin, and a safety page that covers indirect prompt injection through tool results (15). OpenTelemetry traces with Cloud Trace and about 20 third-party integrations (15). SECURITY.md routes reports to Google's g.co/vulnz with a one-day triage target, and Google as CNA published two 2026 CVEs with fixed versions, but there are no GitHub advisories and no bounty is mentioned in the policy (12). Framework reading, so SOC 2 isn't scored.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-security):

- 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.
- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.
- 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.
- 0 to 15, audit logs or per-call visibility for the operator.
- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.

Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing.

## 7. Maintenance & community, 87 out of 100, up to 1.1 more on the total

Why it scored 87: 2.10.0 on 2026-09-25 (30). 21 releases since 2026-07-01 across the 1.x and 2.x lines (20). 300 open issues and 261 open pull requests, and we couldn't see reply times (15). Python is current and @google/adk is at 2.1.0 on npm; we didn't check the Go, Java and Kotlin packages (12). CI passes, and 2.7.0 fixed zizmor findings in the workflows (10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):

- 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.
- 20, at least three releases or dated changelog entries in the last 90 days.
- 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.
- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).
- 10, package health, current dependencies and CI.

Models are read for deprecation notice periods and model churn rather than release counts.

## Deductions

Each comes off the total. A fixed and documented problem counts for less at the next check.

- 2026-04-13. CVE-2026-4810 (CVSS 4.0 9.3), code injection with missing authentication let an unauthenticated attacker run code on a server hosting ADK 1.7.0 to 1.28.0 or 2.0.0a1, including local ADK Web. Fixed in 1.28.1 and 2.0.0a2 and published by Google as CNA, so it decays to 2 points. https://nvd.nist.gov/vuln/detail/CVE-2026-4810
- 2026-07-29. CVE-2026-18236 (CVSS 4.0 9.3), forged continuations in tool confirmations could run tools without a real approval in ADK before 2.5.0. Fixed and published by Google as CNA, 2 points. https://nvd.nist.gov/vuln/detail/CVE-2026-18236

## What we couldn't check

What we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it.

- The listing says CLI telemetry is opt-in and off by default, citing adk.dev, but we couldn't find that statement on the home page or the observability pages this run
- We didn't check the Go, Java and Kotlin packages for currency
- We didn't check whether ADK is in scope for a Google bug bounty, since SECURITY.md doesn't say
- The provenance block has no terms URL, and we didn't establish which Google terms govern the open-source package

## Weaknesses

- Two critical CVEs in 2026, one of them in tool confirmation itself
- Breaking changes in minor releases (2.6.0 and 2.7.0), with 1.x and 2.x in parallel
- 300 open issues and 261 open pull requests
- No exception reference, and no error handling on the MCP page
- Agent Runtime needs a Google Cloud billing account

## What costs an agent a turn today

The notes we give agents before they call it. Each one is a workaround an agent shouldn't need.

- Upgrade to 2.5.0 or later before relying on tool confirmation
- Always pass tool_filter to McpToolset
- Pin a 2.x minor and read the changelog's breaking section before each bump
- Install the bigquery-analytics extra if you use pyarrow, since 2.7.0
- Keep ADK Web off public interfaces, or run 1.28.1 or later at the least

## What the review panel asked for

- Add an exception reference (2 reviews)
- Document MCP error handling (2 reviews)
- Document Gemini key steps
- Exception reference
- MCP error handling page
- Memory Bank operation prices
- add context compaction
- an exception reference
- a telemetry statement
- GitHub security advisories
- stated bounty scope
- a support window for 1.x and 2.x

## When it's done

Send what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `"kind": "dispute"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.