Kagi Assistant
Privacy-focused assistant combines multiple hosted models with optional Kagi Search and user-selected web access.
Score calculation
(4 + 8 + 4 + 1 + 7) ÷ 5 = 4.8
5 of six dimensions rated. Unrated dimensions are excluded from the average.
Full scoring method ↗Setup details
Kagi Assistant: cloud access. Included with Kagi subscription; plan usage limits vary. Ratings describe the stated product, company or configuration in each finding.
What fits your needs
Kagi Assistant subscription service; model selection, web access and limits depend on the selected model and subscription.
Tasks
Combines selected hosted models with optional Kagi Search and user-controlled web access for research and drafting.
See the evidence
Access and billing
Assistant access is bundled with Kagi subscriptions; check current usage allowances and model-specific limits in the account plan.
See the evidence
Tradeoffs to expect
Route
Assistant prompts are sent to the selected model provider through Kagi's service; provider, region and retention details can vary by model and are not fully established by the overview page.
See the evidence
Local setup
The hosted assistant does not itself run models locally; self-hosted model access requires a separate supported endpoint and setup.
See the evidence
How to switch
Get started
Sign in to a Kagi plan with Assistant access, select an available model, and decide whether to enable Kagi Search or web access for each session.
See the evidence
Practical questions
Age and sensitive-use rules
The Kagi Assistant documentation reviewed does not state an age threshold or Assistant-specific sensitive-use boundary. The applicable account terms and model provider policy remain route-specific; no universal age rule is established here.
See the evidence
File and thread controls
Assistant supports document uploads up to 30 MB; file content stays in conversation context. Threads are deleted after 24 hours by default, configurable in settings; users can export/delete threads.
See the evidence
Corrections and provider route
The Assistant lets users switch models and control web access. The reviewed docs do not specify a dedicated factual-error correction route; contact Kagi support for account/product issues.
See the evidence
Kagi Assistant connects users to models from many providers and offers model, Quick profile, web-access, and response-instruction controls. It does not publish request-level energy, emissions, water, or electricity sourcing.
See the evidence
Kagi’s docs explicitly support switching models, choosing cost/speed/context-window tradeoffs, using a Quick profile, disabling web access, and requesting concise answers. These make C2 appropriate. However, the model/provider route is user-selected and no environmental footprint or electricity sourcing is reported; provider/company data cannot be generalized to all Kagi Assistant requests.
1 + 1 reporting + 0 electricity + 2 controls = 4/10. This grade covers documented practices and transparency; it does not measure the energy or emissions of a request.
Upstream model/provider and region per request; request energy, water and emissions; energy used by Kagi routing/search; impact of Quick profile and disabled web access; power sourcing for chosen provider.
- Reporting1/3
- Kagi names supported model providers and lets users change models within a thread, establishing a variable upstream boundary but no environmental quantities.
- Electricity sourcing0/3
- Kagi Assistant can select models across many providers; no route-specific clean-power claim or Kagi-wide AI electricity sourcing is established.
- Resource controls2/3
- Users can select models by speed/cost/context window, use the Quick profile, disable web access, and tailor response-length instructions. These are practical controls, with no measured physical savings.
Water disclosure
Water reporting has not been assessed for this choice.
Kagi offers models from multiple providers and lets users switch models. Its promise not to train on submitted prompts concerns privacy and doesn’t settle the training rights of each provider model.
See the evidence
The exact provider/model is selectable and changes within a thread, so there is no defensible universal training-data rights score for the service. A provider-specific assessment is needed for each model.
The chosen provider and version are required; privacy statements about prompt use do not answer the underlying pretraining-rights question.
Kagi says prompts are not used to train models by Kagi or model vendors and threads delete after 24 hours inactive by default; retention is configurable.
See the evidence
The published no-training commitment and short default thread retention are strong controls; model/provider selection and settings still define actual handling.
Kagi Assistant exposes a selectable catalog spanning proprietary and open-weight families, including OpenAI gpt-oss-120b and Mistral Large 3. Kagi hosts the assistant and doesn’t show that its full service can be self-hosted; each model license and provider route varies.
See the evidence
The model catalog includes genuinely downloadable, modifiable open-weight options, but selecting them through Kagi does not provide the local weights or make Kagi’s search, account and routing stack reproducible.
Kagi’s Assistant documentation lists multiple model providers and lets users configure a thread, but does not report worker conditions, pay, supplier coverage, or independent outcomes for Kagi’s own or upstream model-data labor.
See the evidence
Kagi’s Assistant documentation lists multiple model providers and lets users configure a thread, but does not report worker conditions, pay, supplier coverage, or independent outcomes for Kagi’s own or upstream model-data labor. A provider’s workforce evidence cannot be assigned to all Kagi Assistant routes.
1 + 0 pay & conditions + 0 worker voice & remedy + 0 evidence & follow-through = 1/10. This grades documented worker safeguards, not a measured rate of mistreatment.
Worker pay, exposure and health protections, paid versus volunteer roles, vendor/subcontractor coverage, grievance routes, and independent outcome checks are not established by the reviewed kagi assistant model catalog documentation checked; model selection varies, while workforce and supplier details are not disclosed on that page.
- Pay & conditions0/3
- No opened source reports worker pay, hours, exposure protections, contractor coverage, or worker-level outcomes for Kagi. The reviewed material is limited to Kagi Assistant model catalog documentation checked; model selection varies, while workforce and supplier details are not disclosed on that page.; its silence is an information gap, not evidence of poor conditions.
- Worker voice & remedy0/3
- The reviewed material for Kagi doesn’t show worker representation, grievance access, bargaining voice, or documented remedies. Public code contributions, customer controls, or creator rights do not demonstrate worker voice across employees and suppliers.
- Evidence & follow-through0/3
- No reviewed source provides an independent audit or corroborated worker-level outcomes for Kagi’s relevant employees, contractors, and data-work suppliers. The available disclosures are company/project materials rather than independent labor verification.
Kagi Assistant can route requests to many third-party model providers; its reviewed public product and service terms disclose no Kagi-specific military-use restriction or government deployment.
See the evidence
Kagi’s Assistant documentation describes several third-party models and the ability to change model mid-thread. The reviewed Kagi terms/privacy material describes service operation and fair-use limits, not a military, weapons or surveillance restriction. External providers may apply separate policies depending on model choice; they are not Kagi’s own military relationship or use restrictions. No named government customer or military use was established.
7 base + 0 safeguard adjustment = 7/10 (0 of 9 safeguard points; every 3 adds 1). No military relationship was established in the cited sources. This ranks above known defense ties without claiming a verified clean record.
Current customer contracts, provider-specific enforcement and user-configured deployment patterns remain unknown; route-specific upstream provider policies were not separately scored as Kagi policy.
- Use restrictions0/4
- No Kagi-specific military, weapons or harmful-surveillance AUP found in the opened Kagi product/service materials; upstream provider rules vary by model and are not Kagi’s commitment.
- Enforcement & oversight0/3
- No Kagi military-specific review, remedy or enforcement outcome is disclosed in the reviewed service terms or product page.
- Relationship disclosure0/2
- No Kagi government or defense relationship is established in the reviewed official materials.
The bounded official review found no named military/intelligence contract or operational use for this company/project. This evidence limit is not a clean bill of health; third-party model-provider relationships and customers’ independent deployments are separate.
Military rubric ↗No named target found in the two checked BNC campaign lists.
See the evidence
Company name checked against the BNC consumer boycott/pressure guide and official No Tech campaign list; no named target found. This limited list check is not BDS approval and does not cover every campaign, subsidiary, supplier, or product route.