Kelam

Product Introduction

1. Definition: Kelam is an AI telephony orchestration platform and conversational voice-agent middleware that gives AI assistants, coding agents, and terminal-based workflows a dedicated phone number for inbound and outbound calls. It operates as a phone-number-as-API layer, a persistent two-way voice agent, and an SMS-enabled communication endpoint delivered through a CLI, browser builder, or one-sentence assistant installation.

2. Core Value Proposition: Kelam exists because the real world does not have an API. It lets an AI assistant reach locksmiths, clinics, garages, dentists, rental desks, airlines, tyre shops, salons, and other businesses that have no booking page, no endpoint, no form, and no software integration. By providing a phone number that both dials out and answers inbound calls, Kelam turns voice into the API for everyone who does not have one. The product URL is https://kelam.sh, and it is built by Nousdata Inc., registered in Illinois, USA.

Main Features

1. Two-Way AI Phone Number and Persistent Voice Agent: Kelam provides a workspace with a phone number that remains yours. The same AI agent can dial outbound calls and answer inbound calls on that number, maintaining continuity across conversations. The number can be open to anyone or restricted with an allowlist. If the person you called rings back, the same agent picks up and already knows the previous conversation. This two-way capability distinguishes Kelam from one-way dialers that call and disappear. The inbound demo line is +1 (978) 274-3508, and the agent endpoint supports reads for free while call actions are billed per request or per talk time.

2. One-Sentence Installation and Coding-Agent Integration: Kelam installs with a single instruction: Set up Kelam for yourself — kelam.sh/setup.txt. For terminal users, the CLI installs with curl -fsSL https://kelam.sh | sh && kelam login && kelam skill --install. This writes Kelam’s operating guide into locations that coding agents inspect, including AGENTS.md, Claude Code skills, Cursor rules, and Copilot instructions. Claude Code users can also run /kelam:setup or install the plugin via plugin marketplace add nousdata/plugins and /plugin install kelam@nousdata, which bundles setup and call-data dashboards. Muse, Instinct, Claude, coding agents, and terminal environments can all be given the setup sentence, and assistants with shell access can install the CLI the same way a human would.

3. Outbound Calling, Batch Dialing, and Lead Qualification: Kelam supports outbound calling at scale, including batch jobs such as calling every number in leads.csv, asking whether the contact is still looking, collecting budget information, and returning a CSV with yes/no, budget, and a one-line reason. It can ring three nearest garages to find who can do a clutch on a 2016 Golf this week, what it costs all in, and whether the car must stay overnight. It can call locksmiths near Pioneer Square in Seattle, find one who can arrive within the hour, and return the callout price and ETA rather than a quote form. It can ring rental desks at SFO for a convertible for three days from Friday under $400 all in, check availability, confirm whether debit cards are accepted, and report who is cheapest.

4. Inbound Call Answering, SMS, and Human Handoff: Kelam can answer a shop’s phone line when the owner cannot, collect the caller’s name, number, what they want, and how urgent it is, then text a summary immediately. It can summarise overnight calls, flag complaints, and keep a thread of inbound activity. The same number sends SMS messages, so texts appear in the same thread as calls. For live callers, Kelam can hand the conversation to a colleague, a real phone, or you without dropping the caller. This makes it suitable for clinics, salons, service businesses, and support operations that need voice and text continuity.

5. Phone Tree Navigation and Real-Time Voice Capabilities: Kelam is designed for the queues and menus that software cannot stand in. It waits through hold music, navigates phone trees, works out whether it reached a person, a voicemail box, or a menu, presses the right keys, and leaves a message worth hearing. It can sit through airline hold queues after a cancellation, find out what options are available tonight, and report them before agreeing to anything. The voice agent speaks English, Spanish, and French, and can switch mid-sentence when the caller switches. It supports interruption handling, waiting its turn when interrupted, and maintaining context such as business hours, address, prices, and what the caller wanted last time.

6. Call Recording, Transcripts, Structured Data, and Browser Builder: After a call in either direction, Kelam returns the recording, the transcript, and the facts requested as data. For example, a live call can produce structured fields such as name Maya, booked Sat 2:30, recording, transcript, and text sent. The dashboard at app.kelam.sh/login opens on a chat that can read and write the workspace, and the number comes with it. Users can describe an agent in plain English, such as answering a salon’s phone, booking cuts and colours, asking whether the caller has been in before, texting the address afterwards, and handling price questions. A Call button in the browser lets users speak to the agent with no phone and no cost, then correct it and redeploy in seconds.

7. Cost Controls and Payment Enforcement: Kelam states a hard cap of $2.00 per call bought at call.kelam.sh. The payment challenge enforces the cap technically, not through goodwill. A call that reaches no one, is busy, or fails to connect settles at zero. Users pay for talk time billed to the second. No account is needed to start with the agent endpoint, which takes payment per request with no sign-up and no number to buy. Reads are free. The most lifelike voices are available at cents per minute, billed to the second and capped per call.

Problems Solved

1. Pain Point: The central problem is that the real world lacks APIs. The locksmith, the clinic, the garage, the dentist, the rental desk, and the airline often have no endpoint, no form, and no booking page. Existing AI assistants can reach anything with a URL, but they cannot place a phone call, wait through hold music, navigate a phone tree, leave a voicemail, or receive a callback. Traditional dialers call and disappear, leaving no persistent number for return calls. Manual phone work consumes time, creates inconsistent data capture, and cannot scale for lead qualification, appointment booking, price checking, or urgent service coordination.

2. Target Audience: Kelam is built for AI coding agent users, Claude Code developers, terminal power users, and operators who already use assistants such as Claude, Muse, Instinct, or coding agents. It serves sales and lead-generation teams that need batch calling and CSV outputs, operations managers who need appointment booking and vendor price checks, e-commerce and shop owners who need inbound call answering, customer support teams that need call summaries and complaint flags, executive assistants who need travel and service coordination, healthcare and clinic administrators who need after-hours intake, and founders or marketers who need real-world data collection without an API. It also fits developers who want a CLI-first telephony integration and non-technical users who prefer the browser builder.

3. Use Cases: Specific scenarios include calling locksmiths near Pioneer Square in Seattle to find one available within the hour and return callout price plus ETA; ringing rental desks at SFO for a convertible for three days from Friday under $400 all in and checking debit card acceptance; calling an airline after a cancellation, sitting through hold, and reporting options before agreeing; ringing tyre places around Islington for a 225/45 R17 in stock today, fitted price, and same-day availability; answering a shop’s line, capturing name, number, request, and urgency, then texting a summary; summarising overnight calls and flagging complaints; giving Muse the instruction to ring three nearest garages for a clutch on a 2016 Golf; giving Instinct a Kelam number so it can text and call the dentist on Union Street, book the first opening after 5pm, and text back the result; and running a leads.csv batch to qualify contacts and return structured CSV data.

Unique Advantages

1. Differentiation: Kelam differs from one-way dialers, API-only automation, and assistants that cannot place calls. Muse, Instinct, and Grok Bot launched in summer but none can place a phone call natively. Kelam bridges that gap by giving the assistant a real number that rings both ways. The number stays yours, so a person you called can ring it back and reach the same agent mid-conversation. It handles the queues software cannot stand in, including hold music, phone trees, and rebooking paths that exist nowhere else. It returns recording, transcript, and structured facts as data, making voice conversations usable in downstream workflows. It supports both open access and allowlists, and it can hand live callers to a colleague, a real phone, or the user without dropping them.

2. Key Innovation: The key innovation is treating voice as the API for everyone who does not have one, then packaging that capability into a one-sentence install for AI assistants. Kelam combines telephony, real-time conversational voice, phone tree navigation, multilingual switching, SMS continuity, structured data extraction, and per-call payment enforcement into a single workspace. The CLI writes operating guides into AGENTS.md, Claude Code skills, Cursor rules, and Copilot instructions, so coding agents know how to use the tool. The browser builder lets users describe an agent in plain English, test it in-browser, correct it, and redeploy in seconds. The hard cap of $2.00 per call, second-by-second billing, free failed connections, and free reads create bounded risk. The result is a persistent voice agent that can dial out, answer inbound, text from the same number, and return call data as structured output.

Frequently Asked Questions (FAQ)

1. What is Kelam and how does it give an AI assistant a phone number? Kelam is an AI telephony platform that assigns a phone number to an AI assistant workspace. The assistant can dial outbound calls and answer inbound calls on that number. Installation can be as simple as pasting one sentence into Claude, Muse, a coding agent, or a terminal: Set up Kelam for yourself — kelam.sh/setup.txt. The CLI installs, sends a sign-in link, creates the workspace, and provides the number. The browser option at app.kelam.sh/login also provides a number without a terminal.

2. How do I install Kelam in Claude Code, Cursor, Copilot, Muse, Instinct, or a terminal? For terminal or shell-capable assistants, run curl -fsSL https://kelam.sh | sh && kelam login && kelam skill --install. This writes Kelam’s guide into AGENTS.md, Claude Code skills, Cursor rules, and Copilot instructions. Claude Code users can also add the plugin marketplace with plugin marketplace add nousdata/plugins and install kelam@nousdata. Muse can use its Linux shell and Sentinel approvals to reach kelam.sh and approve the first call. Instinct can be given the Kelam number and briefed on who is on the other end.

3. Can Kelam make outbound calls, answer inbound calls, and send text messages? Yes. Kelam provides a two-way phone number. The same agent can dial outbound, answer inbound, and maintain the same thread. It texts from the same number, so SMS and calls stay together. Inbound callers can reach the agent by calling +1 (978) 274-3508 as a live demo. Outbound use cases include batch calling leads.csv, ringing garages, checking rental cars, waiting on airline hold, and contacting tyre shops. Inbound use cases include answering a shop line, capturing caller details, and texting a summary.

4. How does Kelam handle phone trees, hold music, voicemail, and human handoff? Kelam waits through hold music, navigates phone trees, detects whether it reached a person, a voicemail box, or a menu, presses the correct keys, and leaves a message. It can sit through airline hold queues and report available options before agreeing to anything. For live callers, it can transfer to a colleague, a real phone, or the user without dropping the call. It also knows when to fetch a human, which is critical for urgent, sensitive, or high-value conversations.

5. How much does Kelam cost, and what prevents runaway call charges? Kelam bills by talk time to the second, with a hard cap of $2.00 per call bought at call.kelam.sh. The payment challenge enforces the cap. A call that reaches no one, is busy, or fails to connect settles at zero. Reads are free. The agent endpoint takes payment per request, and no account is needed to start. Voice pricing is cents per minute for the most lifelike voices available, billed to the second and capped per call.

6. What languages does Kelam speak, and can it switch languages mid-call? Kelam speaks English, Spanish, and French. It can switch languages mid-sentence when the caller switches. It also handles interruptions by waiting its turn. The agent retains context such as hours, address, prices, and what the caller wanted last time, so it does not ask twice. This makes it suitable for multilingual clinics, salons, service businesses, and support lines.