Kindred Wildcat — Builder Tools · 2026-10-31 · live
Independent notes. No paid rankings.
01 / The directory
Good tools. Clear trade-offs.
Search by what you want to build. Every listing shows its evidence.
11 listings
Editorial order · not a ranking
No matching tools yet.
Try a broader task or clear your filters to see the full directory.
Coding agentsresearched
Aider
Edit a local Git repository with a chosen model
A researched first pick when explicit file selection and model choice matter more than a visual IDE.
Read the research ＋
Who it suits
Developers comfortable with a terminal and code review
Verdict
A researched first pick when explicit file selection and model choice matter more than a visual IDE.
Trade-off
Model quality and token spend vary; terminal operation is less approachable than a browser builder. Not benchmarked against Cursor or Claude Code.
Privacy and access
Files added to chat are visible to the configured model provider; local CLI does not automatically mean private inference. Review commits and restrict credentials.
Pricing and limits
Checked 2026-10-24: usage docs confirm chosen external or local models and warn that added files cost more tokens. Provider or local inference costs apply; no fixed service price or observed spend established.
A researched first pick for browser-first scaffolding; move security and deployment review outside the prompting loop.
Read the research ＋
Who it suits
Vibe coders who want to start without installing a local toolchain
Verdict
A researched first pick for browser-first scaffolding; move security and deployment review outside the prompting loop.
Trade-off
WebContainer and JavaScript ecosystem constraints; prompt iterations consume tokens. The repository README is not proof of current hosted-service parity.
Privacy and access
Inspect project visibility and hosted-service data terms before adding private source or credentials. An agent can alter the filesystem and run packages.
Pricing and limits
Checked 2026-10-24: Free $0/month with 1M tokens/month, daily limits, public/private projects and Bolt branding. Page displays Pro $18/month equivalent ($216 billed annually), 10M tokens/month with rollover. This is an annual-billing display, not a verified monthly checkout quote; verify promotion and hosting limits before purchase.
Add model-backed chat and tool calls to a TypeScript app
A researched first pick for a common application interface across model providers; it is a toolkit, not a model or free inference service.
Read the research ＋
Who it suits
Builders who can write TypeScript and manage server-side API keys
Verdict
A researched first pick for a common application interface across model providers; it is a toolkit, not a model or free inference service.
Trade-off
Provider features, pricing, limits and retention still differ. Integration and dependency upgrades require tests; no latency or quality comparison performed.
Privacy and access
Keep provider keys server-side; prompts flow to your selected provider. Tool execution needs independent permission controls.
Pricing and limits
Checked 2026-10-24: official introduction describes a TypeScript integration toolkit, not bundled inference. Selected provider or gateway terms apply; no combined project price quote established.
Build customer accounts with relational application data
A researched starting choice when Postgres, authentication and storage belong together.
Read the research ＋
Who it suits
Small SaaS builders prepared to design access policies
Verdict
A researched starting choice when Postgres, authentication and storage belong together.
Trade-off
You still own row-level security, backups and usage controls. Knowledge Base is public beta; migration guides are not an automated migration service or a guarantee against data loss.
Privacy and access
Test isolation with two separate accounts; do not place service-role credentials in browser code. Review hosting region and retention requirements.
Pricing and limits
Checked 2026-10-24: Free $0/month, 500 MB database, 2 active projects, paused after one week of inactivity. Pro from $25/month; additional projects from $10/month and usage overages apply. Guides do not remove hosting or migration costs.
Deploy a commercial web app through a Git workflow
A researched starting choice for a Git-based web deployment workflow, but budget for commercial eligibility rather than assuming Hobby is suitable.
Read the research ＋
Who it suits
Frontend and full-stack builders willing to manage usage budgets
Verdict
A researched starting choice for a Git-based web deployment workflow, but budget for commercial eligibility rather than assuming Hobby is suitable.
Trade-off
Hobby is personal and non-commercial. Pro can incur usage charges; inspect limits and configure budget controls. No hosting benchmark performed.
Privacy and access
Separate preview and production secrets; review logs, regions and access controls before handling customer data.
Pricing and limits
Checked 2026-10-26: Hobby $0/month for personal, non-commercial use. Pro displays $20/month with $20 included usage credit; professional and business use supported. Verify seats, usage charges and checkout before purchase.
Share an interactive result from a cloud coding agent
A researched candidate when sharing an interactive canvas is more useful than pasting a transcript. The October 16 changelog confirms cloud-agent canvases and a beta Environments API. An October 20 entry, recovered October 28, also announces an Agent SDK for code-defined agents, local evals and deployment to Cursor or your own infrastructure; not newly released today.
Read the research ＋
Who it suits
Builders collaborating on cloud-agent work who need a shareable output
Verdict
A researched candidate when sharing an interactive canvas is more useful than pasting a transcript. The October 16 changelog confirms cloud-agent canvases and a beta Environments API. An October 20 entry, recovered October 28, also announces an Agent SDK for code-defined agents, local evals and deployment to Cursor or your own infrastructure; not newly released today.
Trade-off
Cloud execution requires repository-access review and a usage budget. The environments API is beta; maker speed claims have not been benchmarked here.
Privacy and access
Review repository scope, cloud secrets and canvas sharing permissions before sharing customer or proprietary data.
Pricing and limits
Checked 2026-10-28: the official page lists Free $0/month and Premium $30/month with cloud agents and automations; third-party model usage is billed separately. Taxes/fees excluded; verify eligibility and checkout before buying.
Make a bounded repository change while tracking coding-agent usage
A researched candidate for a scoped edit/test/review loop. Use /usage to inspect consumption; do not confuse an API cost estimate with a subscription invoice.
Read the research ＋
Who it suits
Terminal-oriented builders who can review diffs and run project tests
Verdict
A researched candidate for a scoped edit/test/review loop. Use /usage to inspect consumption; do not confuse an API cost estimate with a subscription invoice.
Trade-off
v2.1.314 changes are vendor-reported, not independently tested. The new working-directory write restriction explicitly covers Write, Edit and NotebookEdit; do not mistake it for a complete shell/MCP sandbox. Test permissions and failure handling on a disposable checkout before upgrading.
Privacy and access
Review repository and tool permissions, exclude secrets, and inspect provider data terms before sending private source. Human review remains necessary.
Pricing and limits
Checked 2026-10-30 in official costs docs: API use is token-billed; Pro/Max/Ultra subscribers have included usage subject to plan limits. /usage session dollars estimate API consumption, not subscription billing; managed modelPricing can substitute contracted rates. No fixed subscription price verified; consult linked plan pricing before buying.
Queue and review several local coding agents without losing task history
New, unranked candidate when terminal juggling becomes a problem; start with one read-only task before enabling parallel writes. A queue is not filesystem isolation.
Read the research ＋
Who it suits
Builders already using Claude Code, Codex or OpenCode across projects
Verdict
New, unranked candidate when terminal juggling becomes a problem; start with one read-only task before enabling parallel writes. A queue is not filesystem isolation.
Trade-off
Paid orchestration on top of separate provider costs. Select independent worktrees yourself. Maker reports full validation on macOS; Windows support has narrower validation and Linux is experimental. Not installed or benchmarked here.
Privacy and access
Maker says task history is local and credentials stay with installed CLIs, but prompts reach your model provider. Account/licence checks require connectivity; optional mobile transport has separate boundaries. Local does not mean offline or no data egress.
Pricing and limits
Checked 2026-10-27: 14-day device-bound trial, no card. Solo $7.99 USD/month or $79/year; lifetime $249 once; Team $14.99/seat/month, before tax. Launch offers end January 1, 2027 00:00 UTC; new recurring prices then $9.99/month, $99/year and $19.99/seat/month; lifetime sale ends. Active subscriptions retain starting price. Provider usage/subscriptions extra; checkout not tested.
Extract document facts with source references and a chosen output language
New/unranked candidate when extracted fields need traceable source references rather than an uncited chat answer. Evaluate on redacted sample documents before adoption.
New/unranked candidate when extracted fields need traceable source references rather than an uncited chat answer. Evaluate on redacted sample documents before adoption.
Trade-off
Citations do not prove extraction or translation accuracy. Python >=3.11,<3.15; PDF and DOCX conversion uses an optional Docling extra. Version 0.x integrations require migration. Not installed or benchmarked here.
Privacy and access
Cloud models receive document content; local orchestration does not guarantee private inference. Review model retention and redact sensitive data. Treat document text as untrusted input.
Pricing and limits
Checked 2026-10-28: official README calls the library free and open source; package metadata and LICENSE confirm Apache-2.0. Model-provider charges or local compute costs are separate. No measured extraction cost or paid service quote established.
Build an opt-in browser extension that classifies page text locally
New/unranked candidate for a small local-text feature without a cloud inference backend. Start with Opera One and detect API availability; not a cross-browser production default.
Read the research ＋
Who it suits
JavaScript builders targeting supported Opera desktop users
Verdict
New/unranked candidate for a small local-text feature without a cloud inference backend. Start with Opera One and detect API availability; not a cross-browser production default.
Trade-off
Early preview on Apple-silicon Macs and 64-bit Intel/AMD Windows; not Intel Macs, Linux or Windows on Arm. Official pages disagree on GX/Neon rollout: press release says to follow, blog says available. Installed versions and runtime behavior not tested; hardware affects performance.
Privacy and access
Opera says inference runs on-device, with opt-in model download and site-specific permission. This does not prove a website or extension makes no other network requests. Inspect permissions and network behavior before processing sensitive text.
Pricing and limits
Checked 2026-10-29: official press release describes no account, no fee and no AI subscription for local WebAI use; no per-request cloud inference fee. Device resources and model download are still required. This is not a price claim for Opera Neon or other paid browser services.
Connect an AI assistant to existing app automations without building every integration
New/unranked candidate for a bounded integration prototype. Start with one disposable destination and explicit action approval, not autonomous access to production accounts.
Read the research ＋
Who it suits
Builders prototyping cross-app workflows with supported MCP clients
Verdict
New/unranked candidate for a bounded integration prototype. Start with one disposable destination and explicit action approval, not autonomous access to production accounts.
Trade-off
Credits cover Applet runs, MCP usage and AI features; unlimited Applets does not mean unlimited execution. Per-action credit conversion and overage behavior not established here. Client plan/rollout requirements vary, and free polling is hourly. No authenticated workflow or checkout tested.
Privacy and access
Official setup uses OAuth and connected service accounts; MCP can create automations and execute real actions. Exact OAuth scopes not inspected. Review each service grant, restrict connected accounts and require confirmation before external writes. This is not a read-only connector.
Pricing and limits
Checked 2026-10-30: official USD annual-billing display shows Free $0 with 1,000 credits/month; Starter $2.09/month ($25.08 annually), 5,000 credits; Pro $5.49/month ($65.88 annually), 20,000 credits; Pro+ $13.99/month ($167.88 annually), 100,000 credits. Unlimited Applets on all plans; usage is credit-limited. Monthly checkout and per-action consumption not verified. AI client subscription costs may be separate.
Release dates and discovery dates are distinct. Pricing checks are not announcements.
Editorial week starting 2026-10-26; snapshot through 2026-10-31. This static feed does not refresh itself.
Claude Code — release: v2.1.310 published October 23: Monitor tools extend to Bedrock, Vertex and Foundry; fixes address gateway compatibility, dense-context compaction and Windows hook permission handling. Workflow: use an isolated repository, run a bounded change through your existing gateway, review symlink prompts and hook denial behavior, then compare test results before adopting. Vendor notes only; no gateway or agent execution performed. Recorded 2026-10-26; released 2026-10-23. Source
VibeiDE — new-to-directory: New to this catalog, not a verified new launch: official downloads advertise Desktop 0.8.21; no original release date established. Product guide reviewed October 21 describes per-project queues and completion review. Workflow: select a scratch checkout, ask one installed agent to identify entry point and test command, then verify both before allowing edits. Parallel write tasks need separate worktrees. Recorded 2026-10-27; released date not established. Source
Claude Code — release: v2.1.311 published October 26: fixes premature removal of isolated worktrees used by background subagents, late-connected MCP tools missing from subagents, and stale instruction files after compaction. Benefits builders running multi-agent repository work. Workflow: on a disposable checkout, run a bounded subagent task, check its worktree remains available until completion, then review diffs and tests. Vendor release notes only; not an executed agent evaluation. Recorded 2026-10-28; released 2026-10-26. Source
ContextGem — new-to-directory: New to this catalog; v1.5.0 released October 27, not a new product launch. output_language now accepts BCP 47 language tags beyond en/adapt for string values and justifications. Workflow: use a redacted bilingual agreement, request payment fields and sentence references with output_language='zh-Hans', then compare every value and citation to the original. Maker-documented workflow, not an executed extraction test. Recorded 2026-10-28; released 2026-10-27. Source
Claude Code — release: v2.1.312 published October 27: fixes overwriting large instruction files after partial reads, missing directory-scoped skills after Bash cd, and Code Review processing outdated commits after delayed GitHub events. Benefits repository builders relying on persistent instructions. Workflow: on a disposable checkout, change into a nested directory, verify its scoped instructions load, then inspect the full diff and tests. Vendor release notes only; not hands-on tested. Recorded 2026-10-28; released 2026-10-27. Source
Opera WebAI — new-to-directory / early preview: October 27 press announcement introduces opt-in local Prompt, Decision and Semantic Embeddings capabilities; the developer blog bears October 22, so original earliest availability is uncertain. Workflow: enable Web AI API in Opera One settings, check playground availability, then classify redacted article paragraphs and manually compare labels. Proposed workflow, not an executed vendor test. Use a non-AI fallback for unsupported browsers. Recorded 2026-10-29; released 2026-10-27. Source
Claude Code — release: v2.1.314 published October 30 adds permissions.blockWritesOutsideWorkingDirectories for Write, Edit and NotebookEdit in every permission mode. Also fixes claude -p returning exit 0 with no output on startup failure. Benefits builders running unattended repository tasks. Proposed workflow: on a disposable checkout, enable the restriction, check an outside-path file-tool edit is refused, and assert startup failures exit nonzero. Release-note research, not an executed vendor test. Recorded 2026-10-30; released 2026-10-30. Source
IFTTT.ai MCP — new-to-directory / pricing-check: Official pages now describe IFTTT.ai MCP and credit-based plans. Original release and pricing-transition dates are not established by these undated pages; not presented as a launch today. Proposed workflow: connect a supported client via OAuth to a test IFTTT account, inspect grants, approve a single disposable task action, then verify the destination and credits consumed. Documentation research only; not executed. Recorded 2026-10-30; released date not established. Source
03 / Our approach
Evidence over enthusiasm.
Task-specific starting choices, not a universal or benchmark ranking. Alternatives have not been tested on equal tasks. No numeric scores.
Researched
Reviewed against primary sources. Not a claim of hands-on evaluation.
New / unranked
A discovery worth investigating, not a recommendation or ranked winner.
Hands-on tested
Reserved for a documented vendor evaluation with a task, version and limitations.