AstrBot

# Fix list: AstrBot

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/astrbot, the October 2026 research run, assessed 12 October 2026. Grade C, 58.8 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 AstrBot: 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. Security & auth, 44 out of 100, up to 9.8 more on the total

Why it scored 44: Harness reading of the framework checklist, as the Goose dossier uses. 30 for what leaves the machine by default, 20 for approvals and sandboxing, 15 for prompt-injection posture, 15 for audit and 20 for the security programme. Usage statistics are on by default and go to tickstats.soulter.top with the hostname, an installation UUID, version, model name, adapter name, a hashed sender ID and installed plugin repository URLs. No message text is sent and one variable turns it off (14). No tool call asks for approval, so replies, proactive messages, scheduling and MCP tools run unasked. Shell, Python and file tools are off by default and administrator-only, with a role permission matrix and bubblewrap or Seatbelt isolation on master, and stdio MCP launchers are checked against an allowlist (11). No prompt-injection guidance found. Safe Mode is a content rule in the system prompt (2). Logs, a trace view that is off by default, stored conversations and token statistics (9). No SECURITY.md and no security.txt. GitHub private reporting is on, two advisories are published, one with a CVE, and the repository has a CodeQL workflow. No bounty (8). API keys are scoped and hashed and the panel has TOTP, while the panel binds 0.0.0.0 by default and we found no encryption of stored platform and provider tokens.

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.

## 2. Agent ergonomics, 54 out of 100, up to 7.5 more on the total

Why it scored 54: Harness reading, scored on what an agent or pipeline driving it has to supply. Personas choose tools per MCP server or per tool, a two-stage mode sends tool names before full schemas, and context compression is built in (19 of 25). List routes take page and page size, agent runs stop at 128 steps by default and tool calls have a timeout setting (14 of 20). Responses use one envelope and the scope error is named, with no list of error codes (8 of 20). No idempotency keys or retry guidance, and a one-off scheduled task is deleted after its run even when the run fails (5 of 20). Three commands install and start it, and a chat call needs only `message` and `username`, but there is no client SDK for the HTTP API (8 of 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. Schema & documentation, 62 out of 100, up to 6.2 more on the total

Why it scored 62: Graded on what an outside agent reads, the HTTP API and the built-in tool definitions. The repository publishes an OpenAPI file for `/api/v1` with 170 paths, 240 operations and 60 schemas, and a running server gives it at `/api/v1/openapi.json`. Configuration payloads are open JSON objects by design (22 of 25). No llms.txt on docs.astrbot.app (404) (0). Every operation has a summary and its required scope and no other description. The usage docs say what scheduling does not do and when each MCP transport applies (10 of 20). Built-in tools such as `future_task` have typed parameters with an enum for the action, while `send_message_to_user` names its component types only in prose (9 of 15). One curl example and a table of message segment types. The OpenAPI file documents only 200 responses, and the docs name one error, 403 for a missing scope (6 of 15). Routes carry a `/api/v1` prefix and every release has a dated changelog (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.

## 4. Transparency & trust, 40 out of 100, up to 5.3 more on the total

Made of editorial 56, provenance 23.

Why it scored 40: AGPL-3.0-or-later with the whole source public. The EULA adds conditions on top, among them content restrictions and a grant it calls non-transferable and non-sublicensable, which sits oddly beside the AGPL (26 of 30). The only data statement is section 8 of the EULA. It names no recipient, retention period or field list, and says nothing personally identifiable is collected while the code sends the machine's hostname and an installation ID. There is no privacy policy (8 of 30). Deprecated settings are marked with the version in the config docs and 4.27.0 added runtime warnings for deprecated APIs. No deprecation policy (10 of 20). Telemetry is disclosed in the EULA with an opt-out variable and a config key, is on by default, and its fields are not listed anywhere but the source (12 of 20).

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):

- Legal entity named: not found (0 of 20)
- Domain age: astrbot.app, no registry record we could read (0 of 15)
- Terms of service: read, states 4 of the 7 things a reader expects (7.4 of 10)
- Status page: not found (0 of 10)
- security.txt: not found (0 of 10)

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

Why it scored 60: Read with the self-hosted rule. No payment protocol (0). Free under AGPL-3.0, and the EULA says the team sells no service, so 20, 20 and 20 on the last three lines. It runs with a local model through Ollama or LM Studio and needs no sign-up.

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.

## 6. Reliability, 81 out of 100, up to 3.8 more on the total

Why it scored 81: Read with the local-software lines, as software the owner runs. The README installs `astrbot` with uv on Python 3.12 or later, and a Docker image, a desktop app and Kubernetes files are documented. We read the package name from the README and pyproject.toml and did not open the registry page (20). The unit test workflow runs pytest on Ubuntu, macOS and Windows, with the Windows lane marked best-effort, and the master runs we saw on 12 October 2026 passed, one cancelled (23). 794 open issues and 268 open pull requests on 12 October, with bug reports filed that day on knowledge-base data loss (#10529, #10531) still unanswered and most reports from the previous week showing one or two comments. A stale workflow marks idle bug reports after 60 days and closes them 30 days later (14 of 25). Semantic version tags with betas, and a dated changelog file per release in English and Chinese. No breaking-change section was found in the 4.26 to 4.29 files, though notes name moved settings and the command docs carry an upgrade note (9 of 15). 4.28.2 is a stable 4.x release (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.

## 7. Maintenance & community, 81 out of 100, up to 1.7 more on the total

Why it scored 81: v4.28.2, the latest stable release, is dated 27 September 2026, and v4.29.0-beta.2 followed on 11 October (30). About 15 tags in the 90 days to 12 October 2026 (20). Commits land on master daily and most issues from the past week show a comment, against 794 open issues and 268 open pull requests (15 of 25). Not an MCP server. The PyPI package, Docker image and desktop app are built from the same tags by the release workflow, and we did not open the registry pages (9 of 15). CI passes on master and a CodeQL workflow is present, while Dependabot is configured for GitHub Actions only and dependencies are given as lower bounds (7 of 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.

- 2025-11-14. GHSA-4m32-cjv7-f425 (CVE-2025-55449, critical, 9.8), a hard-coded JWT signing key that let an unauthenticated attacker install a plugin and run commands on any reachable instance at 3.5.17 or earlier. Patched in 3.5.18 on 21 June 2025 and published five months later, inside the last 12 months. Fixed and documented, so the deduction decays to -2. https://github.com/AstrBotDevs/AstrBot/security/advisories/GHSA-4m32-cjv7-f425

## 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.

- Kind. Recorded as `harness`, the closest kind the briefs define for a self-hosted agent program, with `metrics.kind` local. Schema is read on the HTTP API and tool definitions, and Ergonomics and Security with the harness reading the Goose dossier uses
- Category fit. AstrBot is a multi-platform chatbot and agent platform more than a personal assistant. It has scheduled tasks and messaging but no built-in mail, calendar or file-account connectors and no long-term memory of the owner, so `assistant.accounts` and `assistant.memory` are not claimed
- When the hostname field was added to the usage statistics. It is present at v4.28.2 and on master. With no date established we took no deduction and scored it under Transparency and Security
- Whether stored platform tokens and provider keys are encrypted at rest. We found no encryption in the config code and did not run the software
- The advisory for CVE-2025-55449 cites monitoring from AstrBot Cloud of which versions instances run. What AstrBot Cloud is and what it receives was not established beyond the usage statistics and the plugin market at cloud.astrbot.app
- unchecked: the astrbot.app home page, which is script-drawn. Nothing on the listing rests on it
- unchecked: the PyPI and Docker Hub pages. Package names come from the README, pyproject.toml and compose.yml, and release dates from git tags
- unchecked: reply times on individual issues. Only comment counts on the first page of open issues were read
- unchecked: the domain registration date and who operates tickstats.soulter.top
- The docs on master describe 4.29 beta behaviour. The role permission matrix for computer use is not in the 4.28.2 defaults, which have only the runtime switch and the administrator-only flag
- The EULA forbids processing content that involves personal privacy or sensitive personal information, which a personal assistant reading its owner's messages would do. Recorded as a fact, no deduction

## Weaknesses

- No approval step before any tool call. The agent sends messages, creates scheduled tasks and calls MCP tools without asking
- No built-in mail, calendar or file-account connector and no long-term memory of the owner. Both depend on third-party plugins or MCP servers
- Usage statistics are on by default and include the hostname and an installation ID, while the EULA says nothing personally identifiable is collected
- The web panel listens on 0.0.0.0 by default, and the repository has no SECURITY.md or security.txt
- No prompt-injection guidance found, although group messages, web pages and MCP results all reach the model

## 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.

- Set `ASTRBOT_DISABLE_METRICS=1` before the first start. Statistics, including the hostname, are sent to tickstats.soulter.top by default
- Set `dashboard.host` to 127.0.0.1 in `data/cmd_config.json`, or the `DASHBOARD_HOST` variable, unless remote access is intended. The default is 0.0.0.0
- Create an `abk_` key in Settings, OpenAPI with only the scopes needed, and send it as `Authorization: Bearer` or `X-API-Key`
- Include `username` in every `POST /api/v1/chat` body. Without the `chat:admin` scope an administrator ID is rejected
- Check Extensions, Future Tasks after asking for a schedule. One-off tasks are deleted after their run even when the run fails

## 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.