# Reported issues for TypesenseKit

Pod holds 16 of 18 problems reported by people outside the maintainer team. Issues filed by the project's own owners, members and collaborators are excluded entirely — a maintainer's release checklist is not a warning to a prospective user.

Back to [TypesenseKit](/mcp/typesensekit).

## Most discussed

### CLI: confirm destructive operations and support --yes automation

## Problem
Most destructive operation commands execute immediately. Only collection field lifecycle commands have a narrow confirmation mechanism, and there is no consistent interactive safety gate.

## Acceptance criteria
- Destructive operations are centrally classified for CLI safety.
- Interactive terminals receive a clear confirmation prompt naming the operation.
- Non-interactive execution fails safely unless `--yes` is provided.
- `--yes` bypasses the prompt for intentional automation.
- 

[Read the thread](https://github.com/typesensekit/typesensekit/issues/76) · 2026-07-12 · closed · 0 comments

### CLI: add profile test, import, and safe export workflows

## Problem
Profiles cannot be validated before use or moved between machines without manually editing the config file.

## Acceptance criteria
- `profile test [name]` resolves credentials and checks Typesense health.
- `profile export [name]` is secret-safe by default, with an explicit reveal mode.
- `profile import` accepts validated JSON from a file or stdin and refuses accidental overwrite unless requested.
- Imported files may contain one named profile or a complete profile bundle.
- Tests c

[Read the thread](https://github.com/typesensekit/typesensekit/issues/72) · 2026-07-12 · closed · 0 comments

### Make the trusted-publishing workflow idempotent

Changesets can misclassify already-published workspace versions when its npm metadata lookup fails, then attempts duplicate publication and makes an otherwise successful trusted-publishing run fail. This occurred after MCP 1.3.0 published successfully while unchanged CLI 1.2.0 was retried.

Acceptance criteria:
- Preflight each public workspace package against npm before `changeset publish`.
- Temporarily skip versions that already exist without modifying the repository.
- Preserve trusted publi

[Read the thread](https://github.com/typesensekit/typesensekit/issues/66) · 2026-07-12 · closed · 0 comments

### Release TypesenseKit 1.3.0

Prepare and publish the 1.3.0 production MCP release.

Acceptance criteria:
- Consume all pending minor MCP changesets.
- Version `@typesensekit/mcp` to 1.3.0 and update its changelog; leave the unchanged CLI at 1.2.0.
- Full checks, audit, package dry-runs, HTTP tests, and stdio smoke pass.
- Existing Release GitHub Action succeeds through trusted publishing.
- npm reports MCP 1.3.0 as latest and CLI remains 1.2.0.

[Read the thread](https://github.com/typesensekit/typesensekit/issues/64) · 2026-07-12 · closed · 0 comments

### Add secret-safe MCP auditing and transport smoke tests

Production MCP operation needs an audit trail and real transport regression coverage. Resource reads also bypass the tool output redaction path.

Acceptance criteria:
- Add opt-in structured audit events for tool start/success/failure without arguments, results, credentials, or raw error messages.
- Apply secret redaction consistently to resource results.
- Add a real Streamable HTTP MCP initialization test through the secured server.
- Add a spawned stdio client/server smoke test that runs in t

[Read the thread](https://github.com/typesensekit/typesensekit/issues/62) · 2026-07-12 · closed · 0 comments

### Add structured MCP results and tool safety annotations

TypesenseKit MCP tools currently return JSON only as text and do not tell clients whether a tool is read-only, destructive, or idempotent.

Acceptance criteria:
- Return redacted JSON through `structuredContent` while preserving text compatibility.
- Add MCP read-only, destructive, idempotent, and open-world annotations derived from operation behavior.
- Ensure full-mode destructive and raw API tools are accurately marked.
- Add protocol-level tests for listed annotations and structured call res

[Read the thread](https://github.com/typesensekit/typesensekit/issues/60) · 2026-07-12 · closed · 0 comments

### Add MCP execution and response limits

MCP tool calls currently have no process-wide concurrency, rate, timeout, or response-size controls. Large exports or stalled Typesense requests can exhaust an assistant process or hosted server.

Acceptance criteria:
- Enforce configurable process-wide tool concurrency and per-minute rate limits.
- Apply a configurable timeout to every MCP operation.
- Reject serialized tool results above a configurable byte limit.
- Share controls across stateless HTTP server instances and stdio calls.
- Retur

[Read the thread](https://github.com/typesensekit/typesensekit/issues/58) · 2026-07-12 · closed · 0 comments

### Secure the Streamable HTTP MCP trust boundary

The HTTP MCP entrypoint currently binds to all interfaces, accepts arbitrary Origin headers, has no built-in authentication, and buffers unbounded request bodies.

Acceptance criteria:
- Bind to loopback by default with an explicit host override.
- Validate Origin and reject unapproved browser origins with 403.
- Support constant-time bearer-token authentication.
- Refuse unauthenticated non-loopback startup unless an explicit trusted-proxy escape hatch is set.
- Enforce a configurable request b

[Read the thread](https://github.com/typesensekit/typesensekit/issues/56) · 2026-07-12 · closed · 0 comments

## Most recent

### Complete the current Typesense analytics operation surface

TypesenseKit exposes basic rules and event creation, but current Typesense analytics also supports event retrieval, flush, status, rule-tag filtering, and batch rule creation.

Acceptance criteria:
- Add analytics event retrieval, flush, and status operations.
- Support optional rule-tag filtering when listing rules.
- Support current single/batch analytics rule creation while retaining named rule updates.
- Include safe analytics reads in default read-only MCP mode.
- Add route, method, query, 

[Read the thread](https://github.com/typesensekit/typesensekit/issues/50) · 2026-07-12 · closed · 0 comments

### Add first-class Typesense cluster operations

TypesenseKit exposes health, metrics, stats, and debug, but current cluster operations still require raw `api.call`.

Acceptance criteria:
- Add schema-change status, snapshot, leader vote, cache clear, database compaction, and server config operations.
- Only safe operational reads are included in default read-only MCP mode.
- Test HTTP methods, paths, and query parameters.
- Add CLI examples and document destructive/operational behavior.
- Include a minor changeset.

[Read the thread](https://github.com/typesensekit/typesensekit/issues/48) · 2026-07-12 · closed · 0 comments

### Add first-class natural language search model operations

Typesense v30 supports natural-language search models through `/nl_search_models`, but TypesenseKit only exposes conversation models.

Acceptance criteria:
- Add list/create/retrieve/update/delete operations for natural-language search models.
- Include list/retrieve in default read-only MCP mode.
- Test routes, encoding, methods, and request bodies.
- Add CLI examples and root documentation.
- Redact model API keys in returned data and errors through existing safeguards.
- Include a minor chang

[Read the thread](https://github.com/typesensekit/typesensekit/issues/46) · 2026-07-12 · closed · 0 comments

### Add first-class stemming dictionary operations

Typesense supports custom stemming dictionaries, but TypesenseKit currently has no first-class operations for the `/stemming/dictionaries` API.

Acceptance criteria:
- Add list/retrieve operations and JSONL import/upsert for stemming dictionaries.
- Include list/retrieve in default read-only MCP mode.
- Validate and test routes, encoding, query parameters, and request bodies.
- Add CLI examples and root documentation.
- Include a minor changeset.

[Read the thread](https://github.com/typesensekit/typesensekit/issues/44) · 2026-07-12 · closed · 0 comments

### Add first-class Typesense curation set operations

Typesense v30 replaced collection-level overrides with reusable global curation sets, but TypesenseKit only exposes the legacy `overrides.*` surface.

Acceptance criteria:
- Add list/create/retrieve/delete operations for `/curation_sets`.
- Add list/upsert/retrieve/delete operations for curation set items.
- Include read-only list/retrieve operations in default MCP mode.
- Add route, encoding, registry, and read-only tests.
- Document current curation-set support while labeling collection overri

[Read the thread](https://github.com/typesensekit/typesensekit/issues/42) · 2026-07-12 · closed · 0 comments

### Resolve production dependency security advisories

The production dependency graph currently reports two high and nine moderate advisories through the Typesense client and MCP SDK transitive dependency chains.

Affected chains include `typesense -> axios -> form-data` and MCP SDK dependencies on Hono and qs.

Acceptance criteria:
- Upgrade or override vulnerable transitive versions to patched releases.
- `pnpm audit --prod` reports no known vulnerabilities.
- Full checks and package dry-runs pass.
- CI runs a production dependency audit to preve

[Read the thread](https://github.com/typesensekit/typesensekit/issues/36) · 2026-07-12 · closed · 0 comments

### Redact API keys and auth headers from all error output

## Problem

When a network/DNS request failed, the Axios error dump included sensitive headers, including:

```text
'X-TYPESENSE-API-KEY': '...'
```

This is a security bug. CLI errors must not expose API keys or auth headers.

## Expected

Redact secrets in all error output, including nested Axios `config`, `headers`, request objects, response objects, and serialized error causes.

Examples of values that should be redacted:

```text
X-TYPESENSE-API-KEY: [REDACTED]
Authorization: [REDACTED]
api

[Read the thread](https://github.com/typesensekit/typesensekit/issues/11) · 2026-05-22 · closed · 0 comments

### Add first-class collection field add command

## Problem

Adding a single collection field currently requires either the narrow `collections.update` wrapper or generic `tsk api.call`. For vector / embedding fields, `api.call` is required because the field includes attributes like `num_dim`, `vec_dist`, `hnsw_params`, and `embed.model_config`.

## Current workaround

```bash
tsk api.call --input "$(jq -nc --arg openai_key "$OPENAI_API_KEY" '{\n  method: "patch",\n  path: "/collections/production__products",\n  body: {\n    fields: [{\n      

[Read the thread](https://github.com/typesensekit/typesensekit/issues/6) · 2026-05-20 · closed · 0 comments

The remaining reports are on [the project's issue tracker](https://github.com/typesensekit/typesensekit/issues).