Pre-submit Checks
Describe the bug
When connecting to an SSH host through Warp, the session is continuously spammed with:
channel N: open failed: administratively prohibited: open failed
repeating indefinitely with climbing/resetting channel numbers.
Root cause from ps aux: Warp launches ssh as
ssh -o ControlMaster=yes -o ControlPath=~/.ssh/ -t ... export WARP_CLI_AGENT_PROTOCOL_VERSION='1' ... "external_control_master": false ...
Warp creates its own ControlMaster (external_control_master: false) and uses it to open agent-protocol channels. The remote has AllowAgentForwarding no, so every channel Warp opens is refused by sshd, producing the flood.
The problem: this happens even though:
Settings → Agents: agent features disabled
Settings → Warpify → "Warpify SSH Sessions": OFF
Settings → Warpify → "Reuse existing SSH ControlMaster": OFF (Warp still spawns its own master and ignores a user-configured one)
The WARP_CLI_AGENT_PROTOCOL_VERSION=1 injection and ControlMaster=yes fire regardless of these toggles. There appears to be no user-facing way to stop Warp from opening these channels.
To reproduce
Have an SSH host with AllowAgentForwarding no (and/or AllowTcpForwarding no).
Disable Agent features and "Warpify SSH Sessions" in Warp.
ssh from Warp.
Observe continuous channel N: open failed: administratively prohibited.
Expected behavior
With Warpify SSH and Agent disabled, Warp should not open agent-protocol channels, and/or should respect a server that refuses forwarding without retrying endlessly.
Screenshots, videos, and logs
Claude conversation:
The ps aux output showing the spawned ssh command. This is your strongest evidence: it proves ControlMaster=yes, external_control_master: false, and WARP_CLI_AGENT_PROTOCOL_VERSION='1' are injected regardless of your toggles. Paste it as a code block rather than a screenshot so it's searchable/copyable for the maintainer. You can trim the giant bootstrap blob; the diagnostic part is just:
ssh -o ControlMaster=yes -o ControlPath=~/.ssh/1687006307445662789 -t support
export TERM_PROGRAM='WarpTerminal'
export WARP_IS_SSH='1'
export WARP_CLIENT_VERSION='v0.2026.06.17.09.49.stable_02'
export WARP_CLI_AGENT_PROTOCOL_VERSION='1'
hook=... "external_control_master": false ...
[zsh/bash bootstrap truncated for brevity]
Add a line: "(bootstrap body truncated; the relevant flags are ControlMaster=yes, external_control_master: false, WARP_CLI_AGENT_PROTOCOL_VERSION=1)".
A screenshot of the terminal block showing the repeating channel N: open failed: administratively prohibited. The two you already captured are perfect, the first one is ideal because it shows the kubecolor wget succeeding (proving the connection works) immediately above the spam (proving it's cosmetic noise, not a broken session).
The two Warpify settings screenshots proving "Warpify SSH Sessions" is OFF and "Reuse existing ControlMaster" is OFF. These are what make the bug compelling: the toggles say off, the behavior says on.
Server-side confirmation, as a small code block:
$ grep AllowAgentForwarding /etc/ssh/sshd_config
AllowAgentForwarding no
plus the contrast test proving it's Warp-specific:
Clean, no spam:
ssh -o ControlMaster=no -o ControlPath=none oysta@support # silent
macOS Terminal.app: silent
Warp: floods channel-open failures
Operating system (OS)
macOS
Operating system and version
26.5.1 (25F80)
Shell Version
zsh 5.9 (arm64-apple-darwin25.0)
Current Warp version
v0.2026.06.17.09.49.stable_02
Regression
Yes, this bug started recently or with an X Warp version
Recent working Warp date
19th June 2026
Additional context
Only happens in warp
Does this block you from using Warp daily?
Yes, this issue prevents me from using Warp daily.
Is this an issue only in Warp?
Yes, I confirmed that this only happens in Warp, not other terminals.
Warp Internal (ignore): linear-label:b9d78064-c89e-4973-b153-5178a31ee54e
Ignore
Pre-submit Checks
Describe the bug
When connecting to an SSH host through Warp, the session is continuously spammed with:
channel N: open failed: administratively prohibited: open failed
repeating indefinitely with climbing/resetting channel numbers.
Root cause from ps aux: Warp launches ssh as
ssh -o ControlMaster=yes -o ControlPath=~/.ssh/ -t ... export WARP_CLI_AGENT_PROTOCOL_VERSION='1' ... "external_control_master": false ...
Warp creates its own ControlMaster (external_control_master: false) and uses it to open agent-protocol channels. The remote has AllowAgentForwarding no, so every channel Warp opens is refused by sshd, producing the flood.
The problem: this happens even though:
Settings → Agents: agent features disabled
Settings → Warpify → "Warpify SSH Sessions": OFF
Settings → Warpify → "Reuse existing SSH ControlMaster": OFF (Warp still spawns its own master and ignores a user-configured one)
The WARP_CLI_AGENT_PROTOCOL_VERSION=1 injection and ControlMaster=yes fire regardless of these toggles. There appears to be no user-facing way to stop Warp from opening these channels.
To reproduce
Have an SSH host with AllowAgentForwarding no (and/or AllowTcpForwarding no).
Disable Agent features and "Warpify SSH Sessions" in Warp.
ssh from Warp.
Observe continuous channel N: open failed: administratively prohibited.
Expected behavior
With Warpify SSH and Agent disabled, Warp should not open agent-protocol channels, and/or should respect a server that refuses forwarding without retrying endlessly.
Screenshots, videos, and logs
Claude conversation:
The ps aux output showing the spawned ssh command. This is your strongest evidence: it proves ControlMaster=yes, external_control_master: false, and WARP_CLI_AGENT_PROTOCOL_VERSION='1' are injected regardless of your toggles. Paste it as a code block rather than a screenshot so it's searchable/copyable for the maintainer. You can trim the giant bootstrap blob; the diagnostic part is just:
ssh -o ControlMaster=yes -o ControlPath=~/.ssh/1687006307445662789 -t support
export TERM_PROGRAM='WarpTerminal'
export WARP_IS_SSH='1'
export WARP_CLIENT_VERSION='v0.2026.06.17.09.49.stable_02'
export WARP_CLI_AGENT_PROTOCOL_VERSION='1'
hook=... "external_control_master": false ...
[zsh/bash bootstrap truncated for brevity]
Add a line: "(bootstrap body truncated; the relevant flags are ControlMaster=yes, external_control_master: false, WARP_CLI_AGENT_PROTOCOL_VERSION=1)".
A screenshot of the terminal block showing the repeating channel N: open failed: administratively prohibited. The two you already captured are perfect, the first one is ideal because it shows the kubecolor wget succeeding (proving the connection works) immediately above the spam (proving it's cosmetic noise, not a broken session).
The two Warpify settings screenshots proving "Warpify SSH Sessions" is OFF and "Reuse existing ControlMaster" is OFF. These are what make the bug compelling: the toggles say off, the behavior says on.
Server-side confirmation, as a small code block:
$ grep AllowAgentForwarding /etc/ssh/sshd_config
AllowAgentForwarding no
plus the contrast test proving it's Warp-specific:
Clean, no spam:
ssh -o ControlMaster=no -o ControlPath=none oysta@support # silent
macOS Terminal.app: silent
Warp: floods channel-open failures
Operating system (OS)
macOS
Operating system and version
26.5.1 (25F80)
Shell Version
zsh 5.9 (arm64-apple-darwin25.0)
Current Warp version
v0.2026.06.17.09.49.stable_02
Regression
Yes, this bug started recently or with an X Warp version
Recent working Warp date
19th June 2026
Additional context
Only happens in warp
Does this block you from using Warp daily?
Yes, this issue prevents me from using Warp daily.
Is this an issue only in Warp?
Yes, I confirmed that this only happens in Warp, not other terminals.
Warp Internal (ignore): linear-label:b9d78064-c89e-4973-b153-5178a31ee54e
Ignore