AI agent development

An AI agent development company that builds agents you can audit

We build agents that work inside your CRM, calendar, ticketing and email with narrow permissions, a person approving the steps that matter and a log of what they did.

Scope your AI agent
Trusted by the world's fastest-growing brands

What an agent is

An agent acts in your systems, inside limits you set

An AI agent development company builds software that reads a request, decides which steps it needs and carries them out in your CRM, calendar, ticketing tool or inbox. It acts through accounts with narrow permissions, and a person approves the steps that matter.

Agent run · REQ-41279:58 a.m.“Need the mini excavator Fri and Sat.”Look up the customerRead onlyHold the slot, 24 hoursHold onlyConfirm and replyNeeds approvalRefunds and pricesNo access

A chatbot talks

It answers a question in a chat window and stops there. It changes nothing in your systems, so the risk is a wrong answer, and the work is language and accuracy.

A workflow follows a path

When a form arrives, step one runs, then step two. The path is fixed in advance. That makes it predictable and light to run, and it is the right tool for many jobs.

An agent chooses its steps

It reads a messy request and decides whether to look something up, hold a slot, open a ticket or ask a person. Each tool it can call is listed, with what it may and may not do.

Agent or workflow

Two ways to automate a job

Both can read a request and update your systems. They differ in who decides the steps, what happens on an odd input and what each costs to run.

A fixed workflowThe steps are written in advanceAn agent with a narrow jobThe model picks the steps
Who decides the stepsThe builder, in a diagram that runs the same way each timeThe model, from a list of tools it may call
An unusual requestIt follows the closest branch, or fails and alerts someoneIt can reason about it, or hand it to a person
Cost to runFew or no model calls, so cost per run is low and steadySeveral model calls per request, so cost follows the work
How you test itA few sample inputs, since each path is knownA set of realistic requests, scored on what the agent did
SuitsInputs with a clear shape and steps that rarely changeInputs that vary a lot and need judgment between steps
When is a workflow the better buy?

When you can write the steps on one page. A form that creates a CRM contact and sends a confirmation does not need a model choosing what to do, and a workflow costs less to build and to run. We say so in the scope, and quote the workflow. An agent earns its place where the input is open-ended text, the next step depends on what the request says, and a person would otherwise read each one.

What the sources say

Six facts that shape an agent build

Anthropic, OWASP and NIST on what an agent is, where it goes wrong and how to hold it in check. The source sits under each point.

  1. DefinitionAnthropic

    Anthropic separates workflows from agents by who controls the steps.

    Its engineering guide calls workflows systems where models and tools follow “predefined code paths”, and agents systems where the model “dynamically” directs its own process and tool use. We use the same line when we scope a job: you see which steps are fixed and which the agent decides.

  2. Simplest firstAnthropic

    Anthropic recommends the simplest solution that works, which may be no agent at all.

    The guide says this “might mean not building agentic systems at all”, and that agentic systems often trade latency and cost for better task performance. Where the job could be done either way, we say so in the scope.

  3. Excessive agencyOWASP

    OWASP names three causes of damage from an agent: too many functions, too many permissions, too much autonomy.

    Its 2025 entry on excessive agency gives examples: a read-only task connected through an identity that can also update and delete, or an extension that deletes documents without asking the user. The scope lists every tool the agent can call and the permission each one holds.

  4. ApprovalOWASP

    OWASP says to check permissions in the downstream system and to require a person’s approval for high-impact actions.

    The same entry recommends “complete mediation”: your CRM or calendar validates each request against its own rules. It also lists human-in-the-loop approval for high-impact actions. Our agents connect through accounts that hold only the permissions listed, and the approval step is in the written scope.

  5. InjectionOWASP

    OWASP warns that text an agent reads from emails, websites or files can change what it does.

    Its prompt-injection entry describes indirect injection: instructions hidden in content the model reads, such as a web page or a file. It says retrieval and fine-tuning do not fully mitigate it, and lists least privilege and human approval for high-risk actions. A customer’s email is such content, so an agent that acts on it could be steered into a wrong action. The test set includes emails written to try exactly that.

  6. Risk frameworkNIST

    NIST’s AI Risk Management Framework is voluntary and has four functions: Govern, Map, Measure and Manage.

    NIST released version 1.0 on January 26, 2023, and a Generative AI Profile, NIST-AI-600-1, on July 26, 2024. NIST describes it as voluntary guidance. We borrow its vocabulary for the scope: who owns the agent, what it is allowed to do, how it is measured and how it is managed once live.

The scope

What an agent build produces

Six pieces of an agent build, each one a document or setting you can inspect afterwards.

Job definitionDraftjob-definition.pdfApproveAnswer rental requestsHandlesBooking requestsAvailability questionsHands overRefunds and fee waiversPrice changesDone whenA person approves the booking

Job definition

We write down the one job the agent does, the requests it should handle, the ones it hands to a person and what finished looks like. You receive: a one-page job definition, for you to approve.

Action registerridgeline-agentReadCreateEditDeleteLook up customerHold a slot, 24 hOpen damage ticketConfirm and replyRefunds and pricesNo access

Tool and permission list

We list each system the agent touches, each action it can take there and the permission behind it. Reading, drafting and creating are separated from changing and deleting. You receive: an action register, with each permission set in an account in your name.

Approval stepsWaits for a personConfirm bookingREQ-4127RequestExcavator, Oct 9–10ProposesBook and replySourceCalendar holdApproveEdit

Approval steps

We name the actions that wait for a person, who approves them, and what the person sees: the request, what the agent proposes and the source it used. You receive: the approval rules and the screen or message used to approve.

First run: 90 of 100Pass line 96Standard booking requests36 / 38Availability questions18 / 19Changes and cancellations12 / 14Damage, refund and price14 / 15Off-job messages10 / 10CorrectMissedPass line

Test set and pass line

We collect real requests from your inbox, write the correct outcome for each, add messages that try to push the agent off its job, and agree the score that counts as ready. You receive: the test set and the pass line, in your files.

Action logLogging420requestsdays 1 to 30 live10:14 a.m.Booking confirmed, approved10:09 a.m.Damage ticket opened9:57 a.m.Handed to staff9:51 a.m.Question answered

Action log

We record what the agent read, what it decided, what it did and what it handed over, with the time and the request it came from. You receive: access to the log, kept in a system in your name.

Limits sheetFor your approvalcapAt capSteps per requestrun stoppedcapWithinModel spendalert at the capSentry: step cap, REQ-4188to staffcost-per-request.pdfbefore launch

Limits and cost

We set a cap on steps per request, a cap on spend and an alert when either is reached, and estimate model cost per request before launch. You receive: the limits sheet and a cost estimate, for you to approve.

How it runs

Define the job, set the limits, test before launch

Nothing starts until you approve the scope, and you approve each stage before the next one. Each stage hands you a document for your files.

action-register.csv

Action register

Sample · Sample Ridgeline Equipment Rental (Boise, ID)
ActionSystemPermissionApproval
Look up customer and past rentalsCRMRead onlyNone
Check equipment availabilitySchedulingRead onlyNone
Hold a slot for 24 hoursSchedulingCreate a hold onlyNone
Open a damage or complaint ticketTicketingCreate onlyNone
Confirm a booking and send the replyScheduling and emailCreate booking, send emailA person approves
Refund, waive a fee or change a priceBillingNo accessStaff only
Four actions run on their own, one waits for a person and one is outside the agent’s access. The agent connects through its own account, not a staff login.

Our stack

The tools we build and watch agents with

The models, orchestration and monitoring an agent runs on, and the systems it connects to. Select a tool to see what it is and how it fits an agent build.

Anthropic Claude
Why Anthropic Claude

Claude is Anthropic’s family of language models. Through its API, a model can ask for one of the functions you list to be run.

How we use it

We use Claude where the agent reads long or messy requests and chooses between listed tools, and write its data terms into the scope.

ReasoningContextSafety
Anthropic ClaudeStrong
OpenAIStrong
GeminiGood
Open modelsVaries

Questions

AI agent development, answered

Start here

Give the agent one job and see it tested first

Tell us where the agent would work. You get a written scope with the price, and you can decide after reading it.