Merge Agent Handler

# Fix list: Merge Agent Handler

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/merge-agent-handler, the October 2026 research run, assessed 8 October 2026. Grade B, 69.5 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 Merge Agent Handler: 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. Reliability, 59 out of 100, up to 8.2 more on the total

Why it scored 59: Read with the hosted lines. status.merge.dev lists Merge Agent Handler as its own component (20). The page shows 30 days of notices, with one incident in that time, on ADP syncing for the Unified API, and none on Agent Handler. The 60 days before that were not readable, so half marks (15). MCP tool calls are limited per Registered User and per organisation by the minute, with no figures published (3). Third-party limits return `rate_limit_exceeded` with advice to back off, the Free plan's limit returns 402 `billing_limit_reached`, and Registered User creation is idempotent on `origin_user_id`. No Retry-After and no idempotency key for tool calls were found (6). The pricing page lists custom SLAs on Enterprise with no published figure (5). The FAQ says Agent Handler is generally available (10).

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.

## 2. Payments & pricing, 55 out of 100, up to 5.6 more on the total

Why it scored 55: No x402, MPP or L402 (0). Plan prices are public, $1,000 a month for 25,000 credits up to $3,200 for 100,000, at one credit per tool call. The overage rate isn't published, so 15 of 20 (15). The Free plan gives 2,000 credits a month with no card (20). The docs and the OpenAPI file describe an unauthenticated `POST /api/v1/agent/signup/` that returns a key, a tool pack and a registered user. We did not call it, so this rests on the documentation. A paid plan needs a person and a card (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.

## 3. Agent ergonomics, 68 out of 100, up to 5.2 more on the total

Why it scored 68: The catalogue is large (Slack alone lists 50 tools), which scores 5, plus 10 for Tool Packs with per-tool selection, the `connectors` URL scope, the `search_tools` meta-tool and a pin that drops `structuredContent` (15). `tools/list` pages by 100 with `?paginated=true`, and logs and lists take cursors and filters. Controls on tool output size were not found (16). Errors carry an `error_type`, `data.reauth_reason`, a named guardrail on a block and a request id that matches the log row (17). The connectors endpoint returns annotation hints (read only, destructive) per tool. We could not confirm they appear in `tools/list`, and tool calls take no idempotency key (10). There is no Agent Handler SDK. Clients use the MCP SDKs or the Python CLI, and agent signup returns the ids a first call needs (10).

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.

## 4. Schema & documentation, 80 out of 100, up to 3.3 more on the total

Why it scored 80: OpenAPI 3.1 file with 69 operations, and connector tools carry `input_schema` and, where it resolves, `output_schema` (25). llms.txt for the product and a Markdown twin of every page. The OpenAPI link in llms.txt answered 404 and the working file is `agent-handler-api.json` (10). Connector pages give one line per tool, such as 'List all channels in the workspace', with no guidance on when not to use a tool. Tool Packs can override descriptions (12). We could not read live tool schemas without an account. The REST schema types its parameters and has 30 enum uses, while most operations document only a 200 response (10). MCP and curl examples per client and a troubleshooting catalogue with error types and six `reauth_reason` causes (10). The API sits under `/api/v1/`, the MCP protocol revision can be pinned, and the changelog is weekly and dated. No deprecation notices found (13).

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.

## 5. Security & auth, 83 out of 100, up to 3 more on the total

Why it scored 83: Access keys can be limited to named Registered Users and Tool Packs and to the `runtime:all` scope, with expiry, rotation and revocation, and a child key is never wider than its parent. The simplified URL uses OAuth. Two points off because the master key can't be revoked and agent signup returns an unrestricted key (28). Tool Packs select individual tools, input overrides fix field values and a runtime-only key can't change configuration. No confirmation step for destructive tools was found (15). The Security Gateway redacts or blocks tool arguments, and AI Guardrails can screen responses with content passed as data inside boundary markers. Guardrails fail open and new ones screen requests only (11). Tool Call Logs and an Audit Trail are kept 90 days by default with CSV export and an API. The SIEM log endpoints are Enterprise only (14). SOC 2 Type II, ISO 27001, HIPAA, yearly penetration tests and a disclosure programme at security@merge.dev. security.txt returned 404, no paid bounty terms were found, and the programme bars public disclosure without approval (15).

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.

## 6. Maintenance & community, 69 out of 100, up to 2.7 more on the total

Why it scored 69: Changelog entry for Agent Handler on 2 October 2026 and Merge CLI 0.5.0 on PyPI on 5 October (30). Agent Handler entries on 18 September, 25 September and 2 October, and CLI releases on 11 August, 9 September and 5 October (20). A closed service with a weekly changelog, a Discord community on Free, email support on Pro and tickets from the dashboard. The issue tracker linked from PyPI answered 404 (10). No Merge entry was found in the official MCP registry and there is no Agent Handler SDK. The CLI is current (5). The CLI supports Python 3.10 to 3.13 and is classified Beta. Its repository isn't public, so CI was not checked (4).

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.

## 7. Transparency & trust, 71 out of 100, up to 2.5 more on the total

Made of editorial 58, provenance 83.

Why it scored 71: The editorial half. Closed service under a Master Services Agreement that names Merge API, Inc., a Delaware corporation, with California law (15). The MSA says Merge will not use Customer Data to train, fine-tune or improve any AI, and the privacy policy says it doesn't train generalised models on personal information. Logs are kept 90 days by default, a deleted Registered User's records go with it, and a closed account is purged after 30 days. Retention of stored credentials for the life of an account is left to the agreement and the DPA (22). No deprecation policy or dated sunset notices were found. The agreement can change on 30 days' notice or on continued use (4). The sub-processor list, updated 4 May 2026, gives each processor's location and includes Anthropic, OpenAI and AWS Bedrock. Standard plans are hosted on AWS US-East (17).

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: merge.dev, registered 2020-02-09 (6 years) (11 of 15)
- Terms of service: read, states 6 of the 7 things a reader expects, and has 1 clause that costs points (7.1 of 10)
- security.txt: not found (0 of 10)

## 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.

- unchecked: live tool definitions from `tools/list`, including whether annotation hints are present, because reading them needs an account and we did not create one
- unchecked: the unauthenticated agent signup endpoint, which we did not call. Its score rests on the docs and the OpenAPI file
- unchecked: status history older than 30 days, which status.merge.dev did not show us
- unchecked: trust.merge.dev and the Data Processing Addendum text, which we did not read in full
- The numeric rate limits per Registered User and per organisation
- The Pro overage rate per credit, and the Employees Pro price per user
- Whether the Master Services Agreement or the Subscriber Agreement governs a self-serve Agent Handler account. Neither names the product
- Why the repository linked from the CLI's PyPI page answers 404

## Weaknesses

- No rate limit figures are published. The docs say only that MCP tool calls are limited per Registered User and per organisation, by the minute
- Pro starts at $1,000 a month for 25,000 credits, and the overage rate isn't on the pricing page
- Every `tools/call` spends a credit, including failed calls, calls blocked by a rule and calls made with a test key
- AI Guardrails let a call through when the evaluator fails or is unreachable, and new guardrails screen requests only, by default
- The master production key reaches every user and tool pack and can't be revoked, only regenerated. Agent signup returns an unrestricted key

## 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.

- Call `search_tools` with a plain-language intent before `tools/list`. A Tool Pack can hold hundreds of tools and Slack alone lists 50
- Add `?connectors=slack` to the MCP URL to limit a connection to named connectors, and `?paginated=true` only if the client follows `nextCursor`
- Carry a user-scoped key with `runtime:all` only. The production key reaches every Registered User and the management endpoints
- On `reauth_required` read `data.reauth_reason`, then call `authenticate_<connector>` by name. Intent search doesn't return the authenticate tools
- Pin `MCP-Protocol-Version` to `2025-03-26` when context is tight. Later revisions repeat object results in `structuredContent`

## 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.