talkops-ai 0.3.0b13

Ask for any DevOps operation in plain language. TalkOps plans it, runs it safely, and reports back.
Autonomous incident triage · Kubernetes · Terraform · Multi-Cloud · CI/CD — from a local terminal, a pipeline, or a shared server.

Quickstart · Why TalkOps · Smart Mode · Central Server · Webhooks · CLI Reference · Docs

_{Autonomous AWS FinOps Audit in Jev Smart Mode: Installing aws-finops-engineer via /plugins, auditing 17 AWS regions across 32 tool calls (1.1M tokens · 77% cache hit · $0.36 total cost), and generating an actionable remediation report (watch HD MP4).}

What is TalkOps?

TalkOps is an autonomous multi-agent Cloud Operations and DevOps framework built on DeepAgents, LangGraph, LangChain (26+ built-in LLM providers), and the TypeSafe AI Jev System One engine. It runs interactively in a Textual terminal UI (TUI), headlessly inside CI/CD pipelines (talkops -p), over the Agent Client Protocol (talkops --acp), or centrally as a multi-tenant server (talkops serve) that ingests production webhooks and coordinates multiple connected operator clients.

Whether triaging a 3 AM PagerDuty or Prometheus alert storm, diagnosing Kubernetes CrashLoopBackOff pods, auditing multi-cloud AWS/Azure/GCP FinOps and IAM posture, refactoring Terraform modules, or repairing broken GitHub Actions workflows, TalkOps plans and executes end-to-end operational tasks across native Model Context Protocol (MCP) tools and isolated Docker, Kubernetes, Daytona, Modal, and Runloop sandboxes within deterministic safety guardrails.

Quickstart in 60 Seconds

1. Install TalkOps

curl -LsSf https://raw.githubusercontent.com/talkops-ai/talkops/ui-enhancement/scripts/install.sh | bash

# Using uv (recommended for Python 3.12+ environments)
uv tool install --prerelease=allow talkops-ai

# Using pipx
pipx install --pip-args="--pre" talkops-ai

# Using pip
pip install --pre -U talkops-ai

# Install a specific version via install.sh
curl -LsSf https://raw.githubusercontent.com/talkops-ai/talkops/ui-enhancement/scripts/install.sh | bash -s -- 0.3.0b13

2. Run TalkOps (TUI, Headless CLI, or Server Mode)

You can invoke TalkOps using either talkops or the ops shorthand alias:

# 1. Interactive Terminal UI (TUI)
talkops   # or: ops 

# 2. Headless CLI Mode (for shell scripts & CI/CD pipelines)
talkops -p "Audit AWS resources in ap-south-1 for unused volumes and cost savings" 

# 3. Central Server Mode (start the server, then connect from a terminal or CI client)
talkops serve --host 0.0.0.0 --port 2024
talkops --server-url http://localhost:2024 --api-key oc_live_... --user-id alice

For setting up LLM providers, cloud credentials, config.toml, and all other startup options, see the Quickstart Guide, Configuration Guide, and Credentials & Authentication Guide.

Why TalkOps?

Modern engineering doesn't stop at writing code or clicking through a single cloud console. Building and running software spans writing application and infrastructure code, provisioning cloud resources, managing Kubernetes clusters and CI/CD pipelines, monitoring live telemetry, optimizing cloud spend, enforcing security policies, and resolving production incidents.

TalkOps brings coding, cloud management, platform engineering, observability, FinOps, security, and SRE together into a single framework:

How Jev Smart Mode Optimizes Cost, Latency, and Safety

Calling a top-tier reasoning model (claude-opus-5-5, gpt-6-astra, gemini-3.1-pro-preview, deepseek-v4-pro, grok-4.5) with full thinking effort on every status check or webhook event is slow and expensive—yet using a small model on a complex outage investigation misses critical root causes.

Enabling Smart Mode (talkops --smart, /smart, or Shift+Tab) activates the TypeSafe AI Jev classification pipeline. It routes simple checks to fast models in <70ms (cutting token costs by 80–90%), saves heavy reasoning models for deep troubleshooting, and blocks risky production changes in <100ms using zero LLM prompt tokens:

See the Approval Modes & Jev Smart Mode Guide for full details on 3-tier model pools (/pool) and safety thresholds.

Supported Model Providers

TalkOps ships with 26+ built-in model providers out of the box using the provider:model format (for example, talkops -M anthropic:claude-sonnet-5 or /model inside the TUI). You can mix and match providers across your Fast, Standard, and Powerful Smart Mode pools (/pool) or configure custom endpoints in config.toml:

See the Model Providers Guide and talkops auth list for credential environment variables, custom base URLs, and Smart Mode pool configuration.

Centralized Server & Multi-Client Architecture

Instead of running isolated agents on individual laptops, platform teams can deploy talkops serve as a shared TalkOps server that centralizes cloud credentials, model routing, webhook ingestion, and team collaboration behind a single endpoint.

• Shared Investigations & Thread Branching: Multiple engineers can connect concurrently via talkops --server-url http://talkops.internal:2024, browse active or past investigations (talkops threads list or /threads), and branch a live thread (/fork) to test a fix without interrupting teammates.
• Pluggable Server Profiles & Addons: Run the full server (--profile default), isolate API traffic from webhook workers (--profile api-only or --profile webhook-worker, or bind a separate --webhook-port), and install external or custom server addons (talkops addon install or --addons-config addons.json).
• Fast Login & Per-User Privacy: Long-lived API keys (oc_live_...) are exchanged once (talkops auth login) for short-lived signed tokens cached locally with offline grace support, and every thread is automatically scoped to the signed-in user.

See the Central Server & Multi-Client Guide for deployment profiles, server addons, and authentication setup.

Event-Driven Webhooks & Autonomous Incident Triage

When running talkops serve, TalkOps exposes /webhooks/{provider} to receive live production alerts, filter out duplicate alert storms, route follow-up events for the same incident into the same ongoing investigation thread, and automatically deliver Root Cause Analysis (RCA) reports back to your ticketing, source control, and chat platforms.

• Alert Storm Deduplication & Thread Continuity: Repeated alerts for the same incident ID are automatically grouped and routed into the existing investigation thread rather than triggering redundant runs.
• Threaded Chat Updates (Slack, Discord & Microsoft Teams): When an alert arrives, TalkOps can post an immediate "Investigating..." status card to configured Slack, Discord, or Microsoft Teams channels and reply inside that same thread with the completed RCA summary.
• Lifecycle Hooks: Fires webhook.received and webhook.completed events so teams can trigger custom downstream scripts or audit pipelines.

See the Webhooks & Event-Driven Automation Guide for payload formats, signature setup, and write-back configuration.

CLI & Interactive Slash Command Reference

Every TalkOps workflow can be run from the terminal (talkops or ops) or interactively inside the TUI using / slash commands. For the complete flag and command reference, see the CLI & Slash Command Reference.

Core Execution & Server Commands

talkops (Interactive TUI & Headless Execution)

Launches the interactive terminal UI or runs a single task headlessly (-p or piped stdin) for scripts and CI/CD pipelines.

# Interactive TUI in Smart Mode with a specific model
talkops --smart -M claude-sonnet-5

# Interactive TUI in read-only mode (blocks all mutating cloud/K8s/Terraform calls)
talkops --smart --read-only --aws-profile prod-ro

# Headless non-interactive execution with acceptance criteria verification
talkops -p "Audit production EKS ingress controllers for TLS 1.3 compliance" \
  --smart --max-turns 15 --rubric "All ingresses listed with TLS version and remediation diff"

# Connect interactive TUI to a remote Central Server
talkops --server-url http://talkops.internal:2024 --api-key oc_live_... --user-id sre-oncall

talkops serve (Aliases: talkops webhooks, talkops webhook)

Starts the Central Server and inbound webhook listener, inspects active server addons, or lists registered webhook providers.

# Start central server + webhook listener on port 2024 with token auth and Smart Mode
talkops serve --host 0.0.0.0 --port 2024 --auth-mode token --approval-mode smart

# Run with a specific server profile and a dedicated webhook port (8080)
talkops serve --port 2024 --webhook-port 8080 --profile default

# List all 9 built-in webhook provider adapters and their endpoint paths
talkops serve list

# List all discovered built-in and external server addons
talkops serve addons

talkops doctor & talkops help

Checks your local environment (AWS STS identity, installed cloud and DevOps CLIs, Python runtime, and logging) or prints CLI help.

talkops doctor                             # Verify AWS identity, cloud/DevOps CLIs, Python & logs
talkops doctor --aws-profile prod-admin    # Run environment checks against a specific AWS profile
talkops help                               # Show general CLI reference (or `talkops help <subcommand>`)

Environment, Auth & Extensibility Subcommands

talkops auth status                                                    # Inspect session lease, LLM providers & AWS STS status
talkops auth login --sso                                               # Authenticate via Browser SSO (PKCE)
talkops auth login --server-url http://host:2024 --api-key oc_live_... # Exchange an API key for a signed JWT session
talkops auth token --refresh                                           # Print or force-refresh the active bearer JWT
talkops auth list                                                      # List all LLM providers and credential status
talkops auth set anthropic sk-ant-...                                  # Store an encrypted provider API key
talkops auth remove anthropic                                          # Remove a stored provider credential
talkops auth aws --profile prod-admin --region us-east-1               # Verify AWS STS caller identity for a profile/region
talkops auth logout                                                    # Revoke stored session credentials (--all, --account, --provider)

talkops threads list                                                   # List local saved conversation threads
talkops threads list --server-url http://talkops.internal:2024 --user-id alice
talkops threads delete 8f3a9c12-4b2e-4d1a-9e5b-123456789abc            # Delete a thread by ID

talkops mcp list                           # List configured MCP servers
talkops mcp status                         # Probe servers and inspect connection health
talkops mcp probe                          # Connect to configured servers and enumerate live tools
talkops mcp probe --mcp-config ./mcp.json  # Probe servers from a specific MCP config file

talkops plugin list                        # List available and installed plugins
talkops plugin search kubernetes           # Search plugins across registered marketplaces
talkops plugin install eks-pack            # Install a plugin (--scope user|project|local)
talkops plugin enable eks-pack             # Enable an installed plugin
talkops plugin disable eks-pack            # Disable a plugin without uninstalling
talkops plugin uninstall eks-pack          # Remove an installed plugin
talkops plugin refresh                     # Refresh marketplace catalogs
talkops plugin marketplace list            # List, add, or remove plugin marketplace sources

talkops addon list                         # List available and installed server addons (--installed)
talkops addon search pagerduty             # Search server addons across registered marketplaces
talkops addon install custom-ingestor      # Install a server addon and its dependencies
talkops addon enable custom-ingestor       # Enable an installed server addon
talkops addon disable custom-ingestor      # Disable a server addon
talkops addon uninstall custom-ingestor    # Uninstall a server addon
talkops addon refresh                      # Refresh addon marketplace catalogs
talkops addon marketplace list             # List, add, or remove addon marketplace sources

talkops skills list --project              # List discovered skills (omit --project for all scopes)
talkops skills create eks-node-drain       # Scaffold a new SKILL.md runbook (--project for repo scope)
talkops skills info eks-node-drain         # Inspect skill metadata, allowed tools, and path
talkops skills delete eks-node-drain       # Delete a skill directory
talkops skills trust list                  # List, revoke, or clear trusted skill directories

Interactive TUI Slash Commands (/)

Inside the interactive terminal UI, type / to open the fuzzy-searchable command palette. In addition to the built-in commands below, discovered skills (/<skill-name>), plugin subagents (/agent:<name>), and custom commands in .talkops/commands/ are automatically registered in the palette. Every command is documented in the CLI & Slash Command Reference.

Documentation Directory

Explore the complete documentation in docs/talkops-docs/:

Security, Contributing & License

Operational Safety & Security Disclosure

• Fail-Closed Guardrails: TalkOps requires explicit human approval in Smart Mode and Manual Mode whenever an action modifies production resources or carries a high impact risk. Always test custom .talkops/skills/ runbooks, plugins, and MCP servers in a non-production environment or isolated sandbox (docs/talkops-docs/remote-sandboxes.md) before enabling unattended execution in production.
• Untrusted Webhook & Log Inputs: Incoming webhook payloads, log streams, and external repository files enter the agent's context and should be treated as untrusted input. Always enable webhook signature/token verification on talkops serve and scope cloud IAM roles to least privilege.
• Reporting Vulnerabilities: Do not open public GitHub issues for security vulnerabilities. Please follow the private reporting instructions in SECURITY.md.

Contributing

Contributions—including new webhook adapters, SKILL.md runbooks, subagent definitions, MCP integrations, and core framework improvements—are welcome. Review CONTRIBUTING.md and CODE_OF_CONDUCT.md for development environment setup, testing gates, and pull request guidelines.

License

Distributed under the open-source license specified in LICENSE.